Podium presentations are organized into 10 educational tracks. Podium abstracts and speaker information are organized first by track and then by session below.
To search for a specific speaker use the 'Find' functionality in your browser (usually Ctrl + F).
To view a complete schedule of podium presentations and schedule of events for SLAS2020 and to view speaker bios and photos, please visit the SLAS2020 Event Scheduler.
Track Chair(s): Andreas Luippold, Ph.D., Boehringer Ingelheim (Germany) and Martin Giera, Ph.D., Leiden University Medical Center (The Netherlands)
Session Chair: Daniel Bischoff, Ph.D., Boehringer Ingelheim (Germany)
Combining Arrays and Mass Spectrometry for High-Throughput Experiments in Chemistry and Biology
Milan Mrksich, Northwestern University
This talk will describe an approach for using mass spectrometry and arrays of self-assembled monolayers to perform quantitative experiments in high-throughput. The arrays are prepared by immobilizing small molecules, proteins, peptides and carbohydrates to self-assembled monolayers of alkanethiols on gold. These arrays are then treated with reactants—either chemical reagents or enzymes—and then analyzed using the SAMDI technique to identify the masses of substituted alkanethiolates in the monolayer and therefore a broad range of reactivities and post-translational modifications—including kinase, protease, methyltransferase and carbohydrate-directed modifications—and for discovering chemical reactions. This talk will describe applications to high throughput experiments, including the discovery of reactions, the use of carbohydrate arrays to discover novel enzymes, the preparation of peptide arrays to profile the enzyme activities in cell lysates and high-throughput screening to discover novel reactions and small molecular modulators.
These examples illustrate the broad capability of the SAMDI method to profile and discover molecular activities in the molecular sciences.
Establishing Capabilities for (Ultra)High-Throughput Mass Spectrometry at Merck: Reflections from “Year One”
David McLaren, Merck and Co.
Long a prized analytical tool for low and medium-throughput applications, recent innovations in mass spectrometry (MS) are changing the way pharmaceutical scientists apply MS to challenges requiring truly high-throughput. Such endeavors include HTS, parallel medicinal chemistry and protein engineering with emerging needs in drug metabolism and pharmacokinetics as well. At Merck, we have embarked on a journey to incorporate these innovative uHT-MS capabilities in our own approach to drug discovery. This presentation will highlight how we have deployed both a fully-automated uHT-MALDI-MS instrument and the Echo-MS platform in a cross-functional research setting.
As one example, we will demonstrate the application of the uHT-MALDI-MS to a prototypical HTS assay for a kinase target. Here we will discuss how we have enabled "on-demand" MALDI target plate preparation in both 384w and 1536w formats and will further compare & contrast this data to those obtained using a conventional Glo assay for both primary (single concentration) screening and compound titrations.
We will also describe how we have leveraged the Echo-MS platform to triage hits from an HTS designed to discover inhibitors of lipid metabolism which employed a fluorescent, cell-free assay. Here, we will highlight a useful “quench-and-read” sample preparation approach that allows for the selective enrichment of hydrophobic analytes and which is also fully compatible with direct acoustic transfer of the surface solution to the mass spectrometer for analysis.
Throughout the presentation, we will highlight the relevant figures of merit for each application and discuss our findings around the ‘critical parameters’ whose optimization was important to enabling assay success. Finally, we will conclude with a perspective of the future of our journey in this space.
CETSA® Beyond Soluble Targets: A Broad Application to Multi-Pass Transmembrane Proteins
Aarti Kawatkar, AstraZeneca
Demonstration of target binding is a key requirement for understanding the mode of action of new therapeutics. The cellular thermal shift assay (CETSA®) has been introduced as a powerful label-free method to assess target engagement in physiological environments. Here, we present the application of live-cell CETSA® to different classes of integral multi-pass transmembrane proteins using three case studies: the first showing a large and robust stabilization of the outer mitochondrial five-pass transmembrane protein TSPO, the second being a modest stabilization of SERCA2 and the last describing an atypical compound-driven stabilization of the GPCR PAR2. Our data demonstrated that using modified protocols with detergent extraction after the heating step, CETSA® can reliably be applied to several membrane proteins of different complexity. By showing examples with distinct CETSA® behaviors, we aim to provide the scientific community with an overview of different scenarios to expect during CETSA® experiments, especially for challenging, membrane-bound targets.
Novel Label-Free Interaction Technologies and Concepts and its Direct Impact on Early Drug Discovery
Anders Gunnarsson, AstraZeneca
This presentation will disclose three new biophysical concepts and technologies currently implemented in early drug discovery at AstraZeneca to significantly enhance throughput, broaden chemical space or provide an in-depth mechanistic understanding of drug targets and its oligomeric state at the single-molecule level.
Session Chair: Gary Siuzdak, Ph.D., Scripps Research (USA)
Gary Siuzdak, Scripps Research
The metabolome, the small molecule chemical entities involved in metabolism, has traditionally been studied to identify biomarkers in the diagnosis and prediction of disease. However, the value of metabolomics has been redefined from a simple biomarker identification tool to a technology for the discovery of active drivers of biological processes. In this presentation, I will describe the molecular mechanisms by which the active cell metabolome affects cellular physiology through modulation of other "omic" levels, including the genome, epigenome, transcriptome and proteome. This concept of activity screening guided by metabolomics to identify biologically active metabolites or “activity metabolomics”, is having a broad impact on biology.
Connecting High-Throughput Screening and Clinical Pharmacology Using Stable Isotope Tracer Kinetics and Mass Spectrometry: Moving from Enzyme Activity to in vivo Pathway Flux with the Same Assay.
Stephen Previs, Merck
Stable isotope-labeled substrates can be of broad use in cases where target-based high-throughput screening aims to identify compounds that can modulate enzyme activity. For example, depending on the source of a given enzyme target, the presence of endogenous substrates or products can limit one’s ability to follow substrate product conversions; utilization of a labeled substrate(s) can help overcome background contamination. These same isotope flux assays can then be used to follow the progression of hits in later stages of development, including cell-based assays and in vivo studies. Although stable isotope tracer kinetics, coupled with mass spectrometry-based detection, can, therefore, connect all phases of drug discovery some caveats should be recognized to ensure reliable data interpretations.
Our presentation will highlight key areas where the logic surrounding tracer kinetics diverges as the application of flux analyses moves across different stages of drug discovery. We will consider a case study that is focused on lipid biology, i.e. modulating the level of glycosylated ceramides. We will first outline how labeled substrates can be used to circumvent problems that arise in early screening. We will then outline how tracers can be used to progress molecules into later phases, including in vivo studies. Although one can use virtually the same back-end mass spectrometry assay to measure the formation of labeled products, several parameters change with regards to dosing the labeled substrates. For example, when measuring enzyme activity in early biochemical screening one needs to only measure the labeled product. In contrast, in vivo studies must contend with the fact that substantial amounts of “cold” (endogenous) substrate can exist, also, it may not be possible to maintain a steady-state exposure to the labeled substrate. Consequently, strategies need to account for temporal tracer dilution, most of which may not be immediately obvious and/or difficult to correct.
In summary, the ability to measure stable isotope flux from precursors to products can provide a bridge that spans the entire spectrum of drug discovery and development. However, changes in the generation of a labeled product do not immediately reflect changes in the metabolic activity of a given target enzyme or pathway, it is possible to observe differences in the abundance of a labeled product which reflect an unexpected modulation of precursor metabolism. Although the example described here is focused on a targeted screen, the logic has immediate implications with regards to phenotypic screening; attention to a few details can influence essential decision points.
Presentation Title TBD
Baljit Uhbi, Sciex
NMR-Based Metabolomics in Drug Research: Cancer Metabolism
Martin Giera, Leiden University Medical Center
Metabolism and in particular central energy metabolism have evolved as promising drug targets. A cutting edge technology that has become widely used for studying oxygen consumption and extracellular acidification is the Seahorse™ analyzer. While this technology allows for rapid label-free screening it does not provide further details on the involved metabolites and pathways. The quantitative analysis of these pathways, mainly involving glycolysis, the tricarboxylic acid cycle (TCA) and adjacent pathways is intrinsically very challenging. Several commercial solutions have evolved over the years predominantly using mass spectrometry and a series of labeled internal standards. However, many of these approaches suffer from long analytical procedures and the need for special internal standards or kits. As an alternative, we will discuss our NMR based workflow allowing the quantitative analysis of several important pathways for example glycolysis, TCA cycle, OxPhos, one-carbon metabolism and others. Our NMR based workflow allows for the rapid and quantitative analysis of >80 metabolites without the need for specialized kits or internal standards. The workflow can partially be operated in an automated fashion using a KNIME workflow such as KIMBLE. Moreover, flux analysis using 13C labeled materials can easily be adapted resulting in kinetic information.
As an example for the usefulness of this workflow, we will discuss the discovery of choline kinase α (CHKA) as a possible target for the prevention of epithelial to mesenchymal transition (EMT) an important metastatic process. Using NMR based analysis of metabolic changes during TGFα induced EMT we could observe significant alterations in choline phosphorylation. Following up on these results by using experimental inhibitors we could identify CHKA as a crucial enzyme for the EMT phenotype.
Session Chair: Paul Tesar, Ph.D., Case Western Reserve University (USA)
Identifying Compounds that Improve Neuromuscular Function
Lee Rubin, Dept of Stem Cell and Regenerative Biology, Harvard University; Harvard Stem Cell Institute
Motor neuron diseases, as a class, are becoming increasingly well-known and understood. However, effective therapeutics that preserve motor neurons are still lacking. Motor neuron dysfunction is reflected, ultimately, in skeletal muscle weakness and deterioration. Surprisingly, the involvement of muscle cells themselves in diseases ranging from Spinal Muscular Atrophy, a childhood developmental disorder, to sarcopenia – muscle weakness with aging – is much less understood. I will describe two different projects – one directed at promoting the survival of healthy motor neurons, the other at improving muscle health. Both projects started with disease-relevant cell-based screens and culminated in the discovery of new therapeutic targets.
Can Lattice Theory Help Find a Cure for Paralysis?
Nicola Richmond, GlaxoSmithKline
With the advent of the Human Genome Project came the industrialization of the drug discovery process and a belief that combinatorial chemistry and high throughput screening would deliver molecules with increased potency against a single target of interest. Yet the attrition rate is still at the 10% mark and there remain many human diseases for which no effective treatment exists. As Swinney et. al. showed , there is compelling evidence that first-in-class drugs are more likely to be found by assays that measure a clinically meaningful phenotype in a physiologically relevant system rather than a single target-based screening approach in an artificial setting. One perceived issue with phenotypic screening is the lack of mechanistic knowledge. Whilst understanding mechanism of action (MOA) is not a prerequisite for FDA approval, it can guide a medicinal chemistry effort, predict potential toxicities and help define patient populations for clinical trials and ultimately the market place. There are several in vitro approaches to target deconvolution. However, these tend to be of lower throughput and better placed later in a screening cascade. So there is a real need for in silico-based approaches that can be deployed early on in a drug discovery program to identify potential MOAs. Using publicly available data on the Published Kinase Inhibitor Set (PKIS) [2,3], we describe the application of Formal Concept Analysis (FCA), an association mining technique with roots in set theory, to the problem of deconvoluting a phenotypic screen. We describe each compound in the PKIS by the set of kinases it inhibits. We then construct a Galois Lattice, whose nodes correspond to a set of compounds inhibiting a common set of kinases and where two nodes are connected if the compound set of the child node is a subset of the compound set of the parent node. Lattice nodes enriched with compounds that promote neurite outgrowth in rat inform which kinases should be targeted when seeking small molecules that encourage CNS axon repair following injury. The targets we identify using this push-button approach, that can be placed in the hands of the bench scientist, are in line with those identified in  and confirmed in siRNA studies.
1. Swinney DC, Anthony J: How were new medicines discovered? Nat Rev Drug Discov 2011, 10(7):507-19.
2. Drewry DH, Willson TM, Zuercher WJ: Seeding collaborations to advance kinase science with the GSK Published Kinase Inhibitor Set (PKIS). Curr Top Med Chem 2014, 14(3):340-2.
3. Al-Ali H, Lee DH, Danzi MC, Nassif H, Gautam P, Wennerberg K, Zuercher WJ, Drewry DH, Lee JK, Lemmon VP, Bixby JL: Rational Polypharmacology: Systematically Identifying and Engaging Multiple Drug Targets To Promote Axon Growth. ACS Chem Biol 2015, 10(8): 1939-51.
Cytosolic Proteome & Affinity-Based Target Identification (CPATI)
Xianshu Yang, Merck
Disease-relevant phenotypic screening directly identifies ligands that modulate useful biology and constitutes a promising approach to the discovery of novel pharmaceutical treatments. Beyond identifying chemicals with desirable effects, it is important to identify the target(s) and mechanisms that drive the desirable phenotype in complex cellular systems. However, determining the relevant target(s) of phenotypically active ligands has often proven slow or impossible, hampering drug discovery and development. Recently, a few methodologies have emerged that enable detection of target engagement in cells, but most of them require prior chemical modification of either biologically active compounds or proteins. Here, we reported a novel cytosolic proteome & affinity-based target identification platform (CPATI), which is an unbiased, label-free and modification-free approach.
We applied CPATI to identify the candidate target protein(s) of three compounds with Jurkat cells. First, native size-exclusion chromatography (SEC) was used to separate cellular cytosol isolated from Jurkat cell lysis. Second, using affinity-selection technology with our two-dimensional LC-MS system, three compounds were screened with 170 SEC fractions. Cytosolic fractions identified to have specific ligand binding were analyzed via quantitative proteomics. A combination of the protein-bound ligand recovery and target protein SEC elution profile revealed potential targets of test compounds. Third, the thermal shift experiment was conducted to identify proteins with elevated Tm in the presence of ligands, yielding a shortlist of target proteins. Fourth, the target proteins were selected for recombinant protein production and were validated in a binding assay with ligands. Finally, top target proteins were recommended for further validation.
Three control ligands, compound A, compound B, and compound C were identified to bind specifically with Jurkat cytosolic fractions. Their associated target proteins NUDT1 (compound A), HSP90 (compound B), and PAK4 (compound C) were identified from the Jurkat cytosol as their top candidate targets, respectively. Compound A and compound B had a similar binding affinity (Kds) with specific cytosolic fractions and recombinant proteins NUDT1 and HSP90AA1/AB1, respectively. PPP3A-HSP90-CCT complexes were also identified. Compound C is an ATP competitive kinase inhibitor and was identified to associate with seventeen target proteins. Three recombinant proteins PRKACB, PRKCQ and STK38 were confirmed as compound C-bound target proteins.
We also compared thermal shift experiments using compound C with compound C-bound cytosolic fractions and Jurkat crude cytosol. Former samples demonstrated advantage over later samples with short potential target proteins and less false positive. This technology potentially provides a broad application in target and biomarker identification from cells and tissues.
Combining Large-Scale in vitro Pharmacological Profiling and Human Cell-Based Phenotypic Profiling Identifies Novel Mechanisms of Cardiovascular Toxicity
Ellen Berg, Eurofins Discovery
We have previously described a phenotypic signature associated with cardiovascular toxicity relevant to vascular calcification and atherosclerosis from a human primary cell-based coronary artery smooth muscle cell model of vascular inflammation (BioMAP® CASM3C system). The key biomarker activity in this signature is increased cell surface levels of serum amyloid A (SAA) protein. Analysis of a large reference database (BioMAP Phenotypic Profile Reference Database) of >3400 drugs and chemicals tested in this assay identified 147 compounds exhibiting the signature at one or more concentrations. For some of these compounds, specific mechanisms could be implicated and include MEK inhibition, HDAC inhibition, glucocorticoid (GR)/mineralocorticoid (MR) receptor agonism, IL-6 pathway agonism, as well as modulation of mitochondrial NAD+/NADH ratios.
To further characterize the mechanisms underlying this toxicity-associated signature, we took advantage of a second large reference database (BioPrint® Pharmacology Profile Database) comprised of in vitropharmacological profiles of drugs and chemicals screened against a broad range of targets (~148 receptors, ion channels, enzymes and transporters). We evaluated the in vitropharmacology profiles for compounds exhibiting the phenotypic signature associated with cardiovascular toxicity(data was available for 85 of 147 compounds). Target activities (in binding assays) enriched among the phenotypic actives include glucocorticoid receptor (GR), androgen receptor (AR), chloride channel (Cl-channel), ML2 (MT3), (5-Hydroxytryptamine receptor 2B (5-HT2B), peripheral benzodiazepine receptor (BZD), MT1 and ML1. The identification of ML2 (MT3), also known as NAD(P)H quinone dehydrogenase 2 or NQO2, and MT1 receptors is interesting as these are receptors for melatonin. Melatonin has been reported to reduce blood pressure and also to reduce NAD+ levels through effects on NAMPT (nicotinamide phosphoribosyltransferase). Recent studies have suggested that NAMPT may play a role in the pathogenesis of atherosclerosis in experimental mouse models. In humans, serum concentrations of NAMPT are an independent predictor of symptomatic carotid stenosis in patients undergoing carotid endarterectomy.
These results show how the combined analysis of phenotypic and pharmacology profiling data can confirm and extend our understanding of potential mechanisms associated with the risk of cardiovascular toxicity. The pairing of target-based and phenotypic assays is an efficient and effective means to improve confidence in non-animal based screening of new drug leads for potential liabilities.
Track Chair(s): Joe McGivern, Ph.D., Amgen (USA) and Melissa Crisp, Ph.D., Eli Lilly (USA)
Session Chair: James Evans, Ph.D., Phenovista Biosciences (USA)
Presentation Title TBD
Monica Chu, Phenovista
Deep, Single-Cell Analysis by Microscopy: Beyond Human Vision
Anne Carpenter, Broad Institute of Harvard and MIT
Microscopy images contain tremendous information about the state of cells, tissues, and organisms. We aim to go beyond measuring individual phenotypes that biologists already know are relevant to a particular disease. Instead, in a strategy called image-based profiling, we stain many cellular components and extract thousands of morphological features from each cell’s image, often using an assay called Cell Painting. We then harvest similarities in these “profiles” to identify, at a single-cell level, how diseases, drugs, and genes affect cells, which can uncover small molecules’ mechanism of action, discovering disease-associated phenotypes, identify the functional impact of disease-associated alleles and identify novel therapeutics.
Cell-Based Screening to Identify a Lead Humanised Antibody Drug Conjugate
Siobhan Leonard, LifeArc
Antibody-drug conjugates (ADCs) are being designed and used as highly targeted cancer therapies, with five approved by the FDA and over sixty in clinical trials. Building on the success of antibody therapies, ADCs enable highly specific delivery of a toxic payload to a target tumor cell. LifeArc has previous experience in the successful humanization of antibodies for clinical use, Keytruda, Entyvio, Actemra and Tysabri. In order to build upon this expertise and extend the therapeutic approaches, LifeArc has several ADC programmes in the oncology and non-oncology space. As part of this portfolio, the in-house capability has been established and several cell-based assays have been developed to identify candidate ADCs. A case study will be presented that outlines the screening process used to characterize the capacity of candidate antibodies to bind, internalize and induce cell death in HEK293 cells overexpressing the receptor of interest.
Critical to the success of an ADC program is the development of effective methodologies to screen for candidate internalization to the lysosome, where the linker will be cleaved to release the attached cytotoxin. To facilitate high-throughput analysis of hybridoma supernatant and humanized variant internalization, LifeArc has developed robust cell-based assays with the IncuCyte S3 Live-Cell Analysis System and pH-sensitive dyes which fluoresce in acidic lysosomes and endosomes. Following rapid evaluation with the IncuCyte, deeper insight into the intracellular trafficking of promising candidates was gained with the IN Cell Analyzer 6500 HS high content analysis system. When coupled with intuitive analysis workflows to evaluate antibody co-localisation with the lysosome, the sophisticated sensitivity of this high-content imager has allowed further insight into the profile of promising drug candidates and has facilitated the validation of this receptor as an ADC target.
Cell Painting in Hit Discovery
Charles-Hugues Lardeau, AstraZeneca
Having historically looked at characterizing our compound collection for cytotoxicity and cytostaticity or performed the routine annotation of the AstraZeneca compound collection with frequent hitters, we are looking into alternative compound annotation solutions (e.g. imaging or transcriptomics output) and the added benefits they may provide. The Cell Painting imaging assay multiplexes six fluorescent stains to enable the visualization of up to eight cellular compartments in U2OS cells. We will present how we have established the Cell Painting imaging assay in 384-well plates and made it compatible with our CoLab automation platform. The use of a centrifugal plate washer afforded the reduction of staining solution needed and combined with a microvalve-based microplate dispenser allowed the miniaturization of the assay to 1536-well plates. Driven by sustainability, we have also explored the use of a microplate cleaning system to wash and reuse plates for this assay. Specific collections we have applied the Cell Painting imaging assay include a set of compounds also tested with a metabolomics output, compounds that are part of a phenotypic set or compounds from a nuisance compound set. On the analysis side, we will share our experiences using a traditional analysis pipeline (segmentation, working with ca. 1500 features) and the work we are planning to do using deep learning models to best exploit images and data generated. The latter part will touch on the work we have been doing to segment compartments from images.
Session Chair: Amy Quinn, Ph.D., GlaxoSmithKline (USA)
Enzymology Framework and Assay Platform for Targeted Protein Degradation Optimization
Stewart Fisher, C4 Therapeutics, Inc.
Targeted protein degradation, through the use of heterobifunctional degraders that act as catalytic activators for an E3 ligase and target protein, has the potential to transform drug discovery. This talk will discuss the key structure: activity relationships that underpin degrader optimization, including the interpretation and interplay of both thermodynamic and kinetic elements that drive the catalytic turnover of targeted proteins. These data will be placed in context with an enzymology framework to characterize cellular degradation data the extension of these insights to pharmacodynamic modeling and predictions.
Mass Spectrometric Assay of METTL3/METTL14 Methyltransferase Activity
Shane Buker, Accent Therapeutics
A variety of covalent modifications of RNA have been identified and demonstrated to affect RNA processing, stability and translation. Methylation of adenosine at the N6 position (m6A) in mRNA is currently the most well-studied RNA modification and is catalyzed by the RNA methyltransferase complex METTL3/METTL14. Once generated, m6A can modulate mRNA splicing, export, localization, degradation and translation. Although potent and selective inhibitors exist for several members of the Type I S-adenosylmethionine (SAM)-dependent methyltransferase family, no inhibitors have been reported for METTL3/METTL14 to date. To facilitate drug discovery efforts, a sensitive and robust mass spectrometry-based assay for METTL3/METTL14 using self-assembled monolayer desorption/ionization (SAMDI) technology has been developed. The assay uses an 11-nucleotide single-stranded RNA compared to a previously reported 27-nucleotide substrate. IC50 values of mechanism-based inhibitors S-adenosylhomocysteine (SAH) and sinefungin (SFG) are comparable between the SAMDI and radiometric assays that use the same substrate. This work demonstrates the SAMDI technology is amenable to RNA substrates and can be used for high-throughput screening and compound characterization for RNA modifying enzymes.
Fragment-Based Target Screening - An Empirical Approach to Prioritising Targets; A Case Study on Antibacterials
Peter Coombs, LifeArc
In the post-genomic era, abundant functional genomic data is being generated and lists of potential new druggable targets are being analyzed across industry and academia. Bioinformatics and literature diligence can take us so far, but there are two key aspects to target selection that are best examined empirically: tractability and ligandability. We have developed a high-throughput fragment screening approach we call Fragment-Based Target Screening (FBTS) to examine this.
The FBTS approach has been exemplified at LifeArc in its application to antibacterials. New antibacterial drugs are urgently needed to tackle the emerging crisis of multidrug-resistant infections. 50 potential targets were selected by bioinformatic analysis from a TraDIS dataset of essential genes. Our high-throughput expression and purification platform were used to test the targets for expression. Biophysical techniques were used to QC 38 proteins and match them with a control ligand. Proteins were screened against a library of 1280 fragments on a Biacore 8K, then hits confirmed in dose-response experiments. Automated data processing and machine learning were applied to assign a ligandability score to each target & prioritize hit fragments for progression. This was combined with additional essentiality data & a progression assessment to compile a priority list of targets to advance into drug discovery projects.
FBTS is accessible; fragment libraries and the equipment needed for protein production and fragment screening are not only in the domain of big pharma, and the cost and effort associated less daunting than that typically associated with other screening options such as HTS or DNA-encoded libraries. In addition, the approach lends itself to take advantage of the wealth of structural data from public databases and structural genomics consortia. Combined with novel functional genomic data, our empirical tractability and ligandability assessment allows high-quality targets to be prioritized for prosecution and improved chances of success.
Identification and Biophysical Characterization of STING Modulators
Charles Lesburg, Merck & Co. Inc.
The second messenger cyclic dinucleotide (CDN) cGAMP is produced by the cGAS protein in response to activation by cytoplasmic dsDNA. Upon recognition of cGAMP by the stimulator of interferon genes (STING) protein, STING undergoes a substantial conformational change which leads to downstream upregulation of proinflammatory cytokines. Modulation of the cGAS/STING pathway is therefore considered a promising route for the treatment of inflammatory diseases as well as a potential partner for immune-oncology therapies. This presentation will describe the identification and optimization of compounds that were found to antagonize STING signaling as well as the identification and conformational characterization of non-CDN STING agonists. Techniques employed include surface plasmon resonance and X-ray crystallography.
Session Chair: Virneliz Fernandez-Vega, B.S., Scripps Research (USA)
Pancreatic Cancer Patient-Derived Organoids as a Tool for Personalized Medicine
Herve Tiriac, University of California, San Diego
Pancreatic cancer is the most lethal common solid malignancy. Systemic therapies are often ineffective and predictive biomarkers to guide treatment are urgently needed. We generated a pancreatic cancer patient-derived organoid (PDO) library that recapitulates the mutational spectrum and transcriptional subtypes of primary pancreatic cancer. New driver oncogenes were nominated and transcriptomic analyses revealed unique clusters. PDOs exhibited heterogeneous responses to standard-of-care chemotherapeutics and investigational agents. In a case study manner, we find that PDO therapeutic profiles paralleled patient outcomes and that PDOs enable longitudinal assessment of chemo-sensitivity and evaluation of synchronous metastases. We derived organoid-based gene expression signatures of chemo-sensitivity that predicted improved responses for many patients to chemotherapy in both the adjuvant and advanced disease settings. Finally, we nominated alternative treatment strategies for chemo-refractory PDOs using targeted agent therapeutic profiling. We propose that combined molecular and therapeutic profiling of PDOs may predict clinical response and enable prospective therapeutic selection.
Translational Cellular Models
Morag Park, McGill CA
New Innovation to Solve Unmet Needs: Implementing Human Induced Pluripotent Stem Cell-Derived Neural Spheroids as a Robust Screening Platform for Phenotypic-Based Central Nervous System Drug Discovery
Oivin Guicherit, StemoniX
The central nervous system (CNS)-based drug discovery has been hampered by a lack of relevant, high-throughput experimental platforms. Complex, three-dimensional (3D), experimental preparations with multiple cell types better represent the native, in vivo biology, thus providing relevant material for CNS investigations. Unfortunately, these preparations traditionally have not been able to support the throughput necessary for early-stage discovery programs. The ideal preparation would provide consistent native tissue function in high throughput plates. To meet this need, we have developed 96- and 384-well assay-ready, 3D neural spheroid platforms; each spheroid is composed of cortical glutamatergic and GABA-ergic neurons co-cultured with astrocytes to provide a more complex, biologically relevant, and predictive preparation in a high throughput platform for compound screening, safety evaluation, and toxicity studies.
Whole-genome RNAseq profiling demonstrated neural tissue expression patterns, and high content imaging validated neuronal and astrocytic cell populations while showing highly reproducible spheroid size across both 96 and 384-well platforms. Functional neuronal activity was confirmed with MEA recordings and visualized under high-throughput conditions as robust spontaneous, synchronized calcium oscillations with consistent and reproducible baseline activity patterns across wells and plates. Functional circuitry was confirmed by challenging the system with specific ion channel and neurotransmitter receptor agonists and antagonists.
To validate the capabilities of the platform for compound profiling and discovery, a library of 1622 FDA approved compounds were screened in single point at 10 µM final concentration examining Ca2+ oscillations as a functional phenotypic readout. The library included drugs covering a wide spectrum of targets such as CNS biology, oncology, cardiology, anti-inflammatory, immunology, neuropsychiatry and analgesia with DMSO as a vehicle control. Hits were identified as responses that were at least 3 standard deviations from DMSO control responses. As expected, the highest number of hits were from targets associated with neuronal signaling (serotonin, dopamine, GABA, and adrenergic receptors), neural biology, and second messengers such as cAMP. Of note was the identification of several compounds that led to increases in peak count similar to that of 4-AP, a known pro-convulsant. The results validated a robust screening platform with a vehicle control standard deviation of ~9% across all plates and a Z’ score of 0.73 across the entire screen.
In conclusion, performing a high-throughput functional screening assay on our human iPSC-derived 3D neural spheroid platform demonstrated the ability to identify a wide range of hits spanning multiple target areas. This model may serve as a phenotypic and target-based platform for overcoming traditional hurdles of CNS-based drug discovery and improving outcomes for novel CNS-targeted drug discovery and development efforts. Moreover, the model can be created from both wild type and diseased individuals, providing relevant human platforms for disease-specific drug discovery.
Maximizing the Value of Cancer Drug Screening in Multicellular Tumor Spheroid Cultures – Are You Analyzing Your 3D Tumor Models Appropriately?
Paul Johnston, University of Pittsburgh Dept. Pharmaceutical Sci.
Historically, cancer drug leads are identified in high-throughput screening (HTS) growth inhibition assays performed in tumor cell line panels maintained and assayed in 2-dimensional cultures. However, the overall probability for success in oncology clinical trials is a dismal 3.4%. To improve clinical development success rates for solid tumors, more physiologically relevant in vitro 3-dimensional models are being deployed in lead generation to identify better cancer drug candidates. Multicellular tumor spheroids (MCTSs) resemble avascular tumor nodules, micro-metastases or the intervascular regions of large solid tumors concerning morphology, volume growth kinetics, and form diverse microenvironments due to gradients of nutrient distribution and oxygen concentration. Head and neck cancers (HNC) are the 8th leading cause of cancer worldwide and in 2019 it’s projected that 53,000 people in the USA will develop oral cavity or pharynx cancer and 10,860 will die of these cancers. Seven drugs are approved for HNC therapy, but only 10-25% of patients respond to single-agent therapy, and 5-year survival and/or cure rates have not improved. Although pembrolizumab (Keytruda®) was well tolerated in patients with recurrent or metastatic HNC and produced clinically relevant antitumor activity, only 16% of patients responded to treatment. The low response rates and limited efficacy of HNC drugs underscore the need to discover new and effective therapies. We have developed methods to characterize HNC MCTS morphologies, viability and growth phenotypes and to conduct cancer drug HTS. In a total of 95 pairwise cancer drug x HNC cell line experiments, only 35.8% of MCTS cultures exhibited a concentration-dependent growth inhibitory response using metabolic viability reagents, and only 24.4% produced ≥50% reduction in Calcein AM live cell staining. In contrast, 67.8% increased ethidium homodimer dead cell staining by ≥50% and 89.5% altered ≥1 morphological feature; size, shape/perimeter or density/compactness. These data demonstrate that multiple analysis methods are required to accurately assess the impact of cancer drugs on HNC MCTS cultures and to maximize the value of these physiologically relevant tumor cultures.
Session Chair: Dane Mohl, Ph.D., Amgen (USA)
Novel Strategies for Oncoprotein Degradation
Willem Den Besten, Amgen
Targeted protein degradation has the potential to open the door to therapeutic targets previously deemed undruggable. In this talk, I will present the characterization of two ligase ligands and show how target degradation coupled with modulation of ligase biology leads to increased cellular efficacy. I will also share results on a new method for inducing the degradation of an ubiquitin ligase.
Quantitative Live Cellular Assays for Screening Degradation Compounds and their Mechanism of Action
Kristin Riching, Promega Corporation
A new generation of heterobifunctional small molecules, termed PROTACs, holds significant therapeutic potential by inducing degradation of target proteins. These compounds consist of two binding regions separated via a linker: one that specifically binds to the target protein, and the other that directly recruits E3 ligase machinery, resulting in ubiquitination and degradation of the target. Characterizing PROTAC degradation efficacy represents a significant challenge, both in terms of understanding the individual mechanistic processes that control whether degradation will result, as well as the ability to screen for target protein loss in high throughput fashion. Here, we present a live-cell, luminescence-based technology platform that enables characterization and screening of PROTAC compounds and their mechanism of action using either ectopic or endogenous target expression formats. We employ CRISPR/Cas9 endogenous tagging of target proteins with the small peptide, HiBiT, which has a high affinity for and can complement the LgBiT protein to produce NanoBiT luminescence. This allows for sensitive detection of endogenous protein levels in living cells, and can also serve as a BRET energy donor to study protein: protein or protein: small molecule interactions. Using this combinatorial approach, we demonstrate the ability to measure permeability effects and binding affinities of PROTAC compounds to both target and E3 ligase, as well as monitor the kinetics of the subsequent ternary complex (target:PROTAC: E3 ligase) formation, target ubiquitination and recruitment to the proteasome in live cells. We further show the power of this technology in extended kinetic monitoring of endogenous target protein levels, quantification of key degradation parameters for rank-ordering, correlation of these parameters to the precise MOA and the application of these approaches for HTS. This comprehensive technology platform enables rapid, simple and robust screening of functional degrader compounds, ultimately aiding chemical design strategies for the optimization of new therapeutic PROTACs.
Cellular Assays Targeting Two Mutation Classes Causing Cystic Fibrosis: Through (1) Protein Misfolding or (2) Premature Translational Termination
Feng Liang, Cystic Fibrosis Foundation
Cystic fibrosis (CF) is a disease caused by mutations in the gene coding for the cystic fibrosis transmembrane conductance regulator (CFTR), a chloride channel. Mutations are classified into six classes with phenotypes from no CFTR protein synthesis to misfolding and/or functional defects.
Over the past seven years, the FDA approved several novel small molecules that partially correct defects of different mutation classes of CFTR. This has triggered broad efforts to find better and/or different small molecule modulators that address even more CF disease-causing mutations. Here we present screening assays for two classes of CFTR variants: (1) F508del (causing protein misfolding and severely impaired cellular trafficking) and (2) premature termination codon (PTC) mutations, resulting in stop codons in the open reading frame of CFTR and no functional expression. Assays need to address the primary defects of these specific mutation types. A differential screening approach allows the discovery of class-specific hit molecules.
CFTR F508del leads to (1) misfolding of the nucleotide-binding domain 1 (NBD1) of CFTR and (2) perturbs normal interdomain interaction in the CFTR protein. An efficient therapy needs to address both protein folding defects for CFTR for the rescue of CFTR functional expression. Suppressing one defect may allow identification of modulators of the 2nd defect. Thus, using specific suppressor mutations (R555K to restore NBD1 folding or R1070W to rescue domain-domain interactions, allelic screens were developed to enrich for small molecules that preferentially modulate interdomain interactions or NBD1 folding, respectively. The phenotypic screen relies on mammalian cells expressing CFTR F508del with the suppressor mutations and a reporter gene fused into an extracellular loop of CFTR. Hits from the two assays were further tested for complementary effects on the trafficking rescue of CFTR F508del.
A different class of CFTR mutations are PTC variants (about 170 reported) that cannot be treated with available medicines. During CFTR protein synthesis, the interaction of the ribosome with the PTC (UAA, UAG, or UGA) terminates protein translation. Furthermore, when the ribosome stalls at a PTC, translation-coupled RNA surveillance triggers the nonsense-mediated mRNA decay (NMD) pathway, resulting in a reduction of CFTR mRNA levels. Therefore, an effective therapy for CFTR PTC variants needs to address both premature translation termination and reduced CFTR mRNA. Cell-based assays to assess translational readthrough of PTCs have been developed based on either a reporter or the native CFTR gene. RT-qPCR of CFTR mRNA is utilized to monitor anti-NMD effects. Our data support the concept that combining readthrough modulators and NMD inhibitors may lead to more effective therapy.
The CF phenotypes for the above two classes of CFTR mutations derive from defects in different stages of CFTR biogenesis. Specific types of mutations require different screens for the identification of mutation class-specific disease modulators.
Cellular Thermal Shift Assays in High-Throughput: A 1536-Well Cellular Target Engagement Assay for Drug Discovery
Lorena Kallal, GlaxoSmithKline
Thermal shift assays (TSA) reveal changes in protein structure upon binding to small molecules due to a resultant change in the thermal melting temperature of the protein. Experimentally, this change in melt temperature can be measured by exposure of the protein to a temperature gradient, followed by quantification of the protein level or activity at each temperature. Originally, protein thermal shift experiments were performed with purified protein samples, but recently the TSA was reported in a cellular context and the cellular thermal shift assay (CETSA) was born. We have combined CETSA with a high-throughput protein detection method to increase the throughput of the assay since traditional protein detection methods such as western blots are low throughput. To develop high-throughput 1536-well CETSA, we used a protein reporter system in a homogeneous (additions only, no wash) assay format. We have successfully utilized this assay to characterize compounds in dose-response curves for drug discovery programs at GSK. This method can also be applied to identify hits in high throughput screening. Assay parameters optimized included target expression level, the number of detection reagents added after thermal melting, plate type and thermal melt methodologies. Uses and applications in drug discovery will be presented.
Session Chair: Melissa Crisp, Ph.D., Eli Lilly (USA)
Quantitating Endogenous Protein Dynamics with a Bioluminescent Peptide Tag
Marie Schwinn, Promega Corporation
There are an estimated 3,000 human genes that constitute the “druggable genome.” However, only a small percentage of proteins coded by these genes are the focus of drug discovery programs. One barrier in investigating these understudied targets is the lack of easily implemented and scalable methods for assaying proteins. The two principal techniques for analyzing proteins are immuno-detection and mass spectrometry. They offer the advantage of generating data from endogenously expressed proteins. However, these methods are limited by the lack of protein-specific reagents, sensitivity, and HTS compatibility. This prompted us to develop a workflow for studying endogenous proteins that were both easy to use and scalable. In recent years, CRISPR technology has been utilized to integrate reporters into host genomes, such that cellular proteins can be monitored in real-time through detection of the reporter fusion. CRISPR-mediated knockin of the HiBiT luminescent peptide reporter has been demonstrated on a small-scale using a cloning-free workflow. The high sensitivity and dynamic range associated with HiBiT make it suitable to study most cellular proteins across a range of expression levels. Thus, we wanted to determine if CRISPR-mediated HiBiT tagging would provide an approach to rapidly tag any protein in the human proteome. To explore this strategy, a diverse set of proteins representing a broad range of functions and biophysical properties were targeted for tagging with the HiBiT luminescent peptide tag. The majority of the selected targets showed successful integration and expression of functional fusion protein. Given the high success rate in this initial experiment, we investigated if this strategy could be used for developing an HTS-compatible assay for an entire protein family. For this purpose, the cyclin-dependent kinase (CDK) family was targeted for HiBiT tagging and then used to quantitate CDK-specific target engagement. Although the majority of edited CDK-HiBiT cell lines displayed compound pharmacology similar to what was observed in over-expression-based models, several differences were found which suggests that endogenous models may provide more accurate information on compound activity. In summary, CRISPR-mediated tagging of endogenous proteins with HiBiT represents an easy and scalable strategy for studying endogenous proteins which enables the analysis of proteins in their appropriate physiological context.
A 384-Well Workflow to Execute an Arrayed CRISPR-Cas9 Gene Editing Screen in T-Cells
Sapna Desai, GlaxoSmithKline
Functional genomics approaches to identify novel therapeutic targets are rapidly gaining traction. Arrayed screening for the phenotypes resulting from gene-knockouts using CRISPR-Cas9 technology can yield results rapidly, with very little need for target deconvolution. Data can be further enhanced by the selection of disease-relevant primary cells.
We have developed a high-efficiency, arrayed genome-editing screen in primary CD4+ T cells using CRISPR–Cas9 for the identification of genes associated with cytokine release. T-cells are isolated, purified and expanded before genome editing occurs via nucleofection. A 384-well nucleofector is used to deliver RNP complexes consisting of guide RNA (gRNA), transactivating CRISPR RNA (tracrRNA) and Cas9 enzyme. Edited cells are rested and activated before being utilized in downstream assays capturing multi-cytokine release and cell viability.
The development of miniaturized, robust nucleofection protocols and assays for T-cell screening allows integration of this challenging cell-type onto well-established liquid handling platforms and demonstrates the potential of genome-wide arrayed CRISPR-Cas9 screening of primary cells in a screening environment.
Development and Implementation of High-Throughput Cellular Protein Stability Assays for Evaluation of Target Engagement at Early Stages of Screening Projects
John Holleran, Sanford Burnham Prebys Medical Discovery Institute
Cellular protein thermal stability provides a powerful method for assessing compound target engagement. Recently, there have been several high-throughput 384 well assay compatible detection formats published for cellular protein thermal stability using commercial luminescence complementation systems as well as homogenous antibody sandwich AlphaLISA assays. Similarly, we have utilized these detection formats and successfully developed and employed high-throughput format assays for a large number of diverse drug discovery projects ongoing at the SBP Prebys Center. In addition to serving the primary goal of assessing the target engagement, these efforts have provided in-depth knowledge of diverse detection systems, such as Promega HiBiT, DiscoveRx ePL, AlphaLISA and the classical approach relying on SDS-PAGE-Western Blot detection. In addition to monitoring the thermal stability of intracellular proteins, we successfully employed these same detection approaches to assess small-molecule effects on steady-state protein level or localization in the cell. Not only do these approaches enable reporting on direct target binding, but they also monitor the effects of small molecules on the protein target interactome governing homeostasis. This enables the identification of druggable partners from the entire target interaction network. Along with small molecule screening, all these methods provide unique and powerful tools to study targets of interest in their native state which exposes valuable information about the effects of cellular background and extracellular environment. From these studies, we have gleaned insight into the cellular and context-specific target regulation and potential biological relevance of identified hits. We have successfully utilized both exogenous expression and CRISPR knock-in of detection tracers or AlphaLISA detection to assess protein stability of endogenous proteins learning advantages and disadvantages of each approach. Side-by-side assessment of these approaches helped to develop a decision tree for selecting the most appropriate approach for each new project and target.
The Development and Application of a Whole Genome Arrayed CRISPR Screening Platform for Target Discovery and Mechanism of Action Investigation
Douglas Ross-Thriepland, AstraZeneca
The identification of novel therapeutic targets that translate into clinical successes is needed now more than ever to deliver life-changing medicines to patients across disease areas; from drug resistance in oncology to cardiovascular and respiratory disease. In this effort, the unbiased identification of targets through perturbation at the gene level is not new. However, the CRISPR/Cas9 revolution has enabled us to achieve this with higher efficiency, reduced off-target effects and has enabled new modes of perturbation such as gene activation (CRISPRa) and SNP mutation (base editing). We have used this technology to build a Target Discovery platform encompassing both pooled and arrayed screening techniques that, taken together, allow us to probe a wide range of biology, from the slow onset of drug resistance seen in oncology (pooled), through to understanding what role genes play in signaling and cell response (arrayed). Here we present the development and application of our Arrayed CRISPR Screening component of this platform. Using CRISPR libraries comprised of synthetic gRNAs arrayed into a “one-gene-per-well” format we demonstrate the high-efficiency of both genes knock out (CRISPRn) and gene activation (CRISPRa) at whole genome-scale in cell-based assays. Coupled with the generation of high-quality Cas9 expressing cells through the ObLiGaRe insertion of an inducible Cas9 expression cassettes (ODin), we show that CRISPR/Cas9 is a powerful and robust technology for arrayed screening. By combining this platform with high-content imaging and multivariate analysis technology we have been able to interrogate the phenotype resulting from gene perturbation to a much greater depth. This has significantly improved how we are able to rank, triage and progress hits into target validation and into the clinic. Here we demonstrate the application of this platform for the unbiased discovery of new therapeutic targets with two case studies that exemplify the capability and how it has impacted our new target pipeline in both oncology and advanced drug delivery.
Session Chair: Julie Conkright-Fincham, Ph.D., Stowers Institute for Medical Research (USA)
Functional Proteome Array Screening Strategies for Biomarker Discovery
Joshua LaBaer, The Biodesign Institute, ASU
Self-assembling protein microarrays can be used to study protein-protein interactions, protein-drug interactions, search for enzyme substrates and as tools to search for disease biomarkers. In particular, recent experiments have focused on using these protein microarrays to search for antibody responses in patients with cancer, autoimmune and infectious diseases. This approach has led to the first CLIA-certified blood test for the early detection of breast cancer, Videssa™. Recent work has focused on using the arrays to explore the post-translational modification of proteins and their role in producing neoantigens in disease.
NanoClick Assay: A High-Throughput, Target-Agnostic Cell Permeability Assay that Combines NanoBRET Technology with Intracellular Click Chemistry
Andrea Peier, Merck
Macrocyclic peptides open new opportunities to target intracellular protein-protein interactions (PPIs) that are often considered non-druggable by traditional small molecules. Specifically, peptides have the potential to bind to highly expansive binding surfaces (orthosteric blocking) of such PPIs and/or other unique allosteric binding sites. However, their clinical development may be limited by their ability to efficiently penetrate cells to modulate their cognate PPI targets. The ability to have a predictive, high-throughput assay to assess cell permeability is a critical tool to support peptide drug discovery programs.
We developed a high throughput, quantitative, target-agnostic cell permeability assay that essentially measures the cumulative cytosolic exposure of a peptide in a concentration-dependent manner. The assay has been named NanoClick as it combines in-cell Click chemistry and monitoring of a NanoBRET signal in cells. The assay is based on cellular expression of the NanoLuc-HaloTag system and relies on the Click reaction of azide-containing peptides with DiBac-chloroalkane (CA) anchored to the HaloTag. The subsequent introduction of an azido-dye followed by the NanoLuc substrate allows the detection of a BRET signal that is reduced by the presence of Click-reactive peptides in the cytosol. The readout can be expressed as a permeability ratio of EC50s when compared to the response of a low permeability control.
We validated the assay using known cell-penetrating peptides and were further able to demonstrate correlations to cellular activity using a p53/MDM2 model system. The assay has been applied across multiple programs and has been used to guide and establish structure-permeability relationships in the optimization of macrocyclic peptides for cellular potency across intracellular PPI target programs.
Building Toolkits for the Orphan Kinome
Laurie Parker, University of Minnesota
Protein phosphorylation by kinases is a major mechanism of cell signaling and is involved in almost all aspects of cell biology. Kinase dysregulation is a key factor in diseases like cancer, and kinases are one of the major drug targets in oncology. However, despite decades of research and billions of dollars in drug discovery efforts on kinases, relatively few are well characterized. The majority of the ~90 tyrosine kinases are considered “orphans,” for which few to no substrates, and thus few details about biological pathways and roles, are known. Without substrates to use as activity probes, inhibitors for use as tool compounds or potential therapeutics cannot be discovered. We have developed a strategy to incorporate empirically-determined substrate profiling data into our KINATEST-ID bioinformatics pipeline to efficiently tackle the orphan kinase problem, determine substrate preferences and design novel substrate tools. Protease-digested peptides from cell lysates are stripped of pre-existing phosphates, then re-phosphorylated with a kinase of interest. The resulting phosphopeptides are enriched and analyzed using mass spectrometry. Phosphopeptide sequences are extracted from the peptide ID list and funneled through the KINATEST-ID pipeline using a set of scripts implemented in the open-source user interface GalaxyP, to define substrate sequence preferences and propose candidate optimal substrate peptides. Those are then synthesized and tested for phosphorylation efficiency by the target kinase. Using this approach, we have characterized substrate preferences for several understudied kinases for which few validated substrates were known, including FLT3 and two clinically relevant mutants, and BTK. Current and future efforts are to broaden the scope of kinases characterized using this streamlined phosphoproteomics/bioinformatics pipeline and proceed with systematically defining substrate information and developing novel tools for other orphaned kinases in the kinome.
Hybridization Chain Reaction for Single-Cell Visualization of RNA in High-Content Imaging Assays
Gianluca Pegoraro, National Cancer Institute
The precise regulation of gene expression programs is responsible for the establishment and maintenance of cell, tissue and organ identity, for cellular responses to signaling cues and injuries, and, when disrupted or rewired, for diseases such as cancer and inflammation. Measuring gene expression in high-throughput assays often requires reporter cell line engineering, or using antibodies against endogenous protein markers, which involves a lengthy development process, and can also suffer from batch-to-batch variation. On the other hand, single-molecule RNA Fluorescence In Situ Hybridization (smRNA-FISH) detects endogenous transcripts, and is based on DNA oligonucleotide probes that can be rapidly designed in silico, chemically synthesized, tested, and scaled up. For these reasons, smRNA-FISH has the potential to be a useful additional tool for High-Content Imaging (HCI) in chemical or functional genetics screens for the identification of gene expression regulatory pathways. However, visualization of RNA at the single-cell level via smRNA-FISH has not been optimized for HCI assays. To address these limitations, we adapted the single-step, enzyme-free RNA Hybridization Chain Reaction (RNA HCR) to a 384-well format using an HCI platform. First, we used RNA HCR probes against IFIT3, an interferon-stimulated gene (ISG), to demonstrate that high-throughput RNA HCR can quantitatively measure gene expression changes at the single-cell level in a 384-well format. As a proof of principle, we performed a focused RNAi screen against 521 human genes involved in epigenetics regulation to identify novel factors mediating the transcriptional response to interferon-γ. The results of this primary screen suggest that multiple components of the MOF acetylase complex are involved in the upregulation of IFIT3 upon interferon stimulation. Finally, we applied high-throughput RNA HCR in other HCI assays to measure expression levels of specific mRNA splicing isoforms of the FGFR2 gene, to monitor the effect of steroid treatment on the expression of inflammation regulators in primary human monocytes, and to determine the effect of steroid treatment on a variety of GR-responsive genes in mouse cells. Altogether, these results indicate that RNA HCR can be miniaturized in 384-well assays to semi-quantitatively detect several endogenous RNA species via HCI in physiologically relevant systems, at the single-cell level, and in a medium- to high-throughput format. In the future, we expect that high-throughput RNA HCR will be useful for the discovery and validation of diverse targets regulating gene expression.
Session Chair: James Pilling, M.Sc., AstraZeneca (UK)
Revolutionizing Cellular Screening with Artificial Intelligence-Driven Label-Free Imaging
Adam Corrigan, AstraZeneca
The primary reasons for drug failure in the clinic are a lack of efficacy and safety. Therefore, to drive a better understanding of disease biology and improve the translation, cellular imaging assays in early discovery need to be increasingly complex, utilizing multiple biomarkers to label several proteins in a pathway and to quantify multiple sub-populations.
The field of image analysis has been transformed by the explosion of machine learning and AI methods, and we are now leveraging recent developments to maximize the information we get from imaging data and enable new experimental approaches. A key limitation of machine learning, and particularly deep learning models is the requirement for large amounts of annotated training data. We have developed an active learning framework for efficient training data generation, alongside unsupervised phenotype discovery approaches, to build models that can quantify the full complexity of cellular screening data.
We are also integrating label-free phase-contrast imaging into our cellular screens. A large amount of information on cellular morphology is contained in the phase-contrast images, which do not take up a fluorescent color channel, but the human interpretation is very difficult. By training a deep neural network to find and segment nuclei and cells from phase-contrast images alone, nuclear and cell markers are no longer required. This allows multiple biomarkers to be combined into a single screen, enabling more complex biology for less cost.
In addition to segmentation, we have shown that standard readouts such as cell division and cell death can be predicted from the label-free images, opening the possibility for digital multiplexing of a wide range of biomarkers, in living cells, and without expensive cell engineering. Combining this method with the high-content Cell Painting assay, we are learning how to extract meaningful biological features from label-free images, which can then be used to re-interrogate existing screening data for new insights. This approach is straightforwardly integrated with existing workflows and is revolutionizing the questions being asked through cellular screening.
Applications of Image-Based Artificial Intelligence in Drug Discovery and Safety Testing
Mahnaz Maddah, Dana Solutions LLC
Discovering effective drugs and demonstrating their safety are significant challenges facing the pharmaceutical industry, due to the high costs of development, long lead times, and low success rates of late-stage clinical trials. There is a need for new tools and technologies to help identify safe and effective drugs during the early stages of development.
Over the last decade, there has been significant progress in using human induced pluripotent stem cells (hiPSCs) for modeling of human disease, drug screening, and toxicity testing. Numerous studies have demonstrated that these cells have physiologically relevant characteristics and can be used for preclinical testing of new drugs using high-throughput assays. In such assays, image and signal analysis algorithms are used to generate quantitative measurements that relate to cell degradation, death, or changes in function. Such approaches may be missing subtle changes that are not easily visualized, are too complex to measure with traditional data analysis methods, and/or suffer from lack of consistent quality control metrics on the input data.
Artificial intelligence (AI) techniques and specifically deep convolutional neural networks are perfectly suited to address the challenges of these high-throughput assays by analyzing large amounts of imaging data robustly and with a level of sensitivity that has not been previously possible. We present case studies for using AI for high-throughput image-based phenotypic screening, toxicity testing, and quality control. First, we present data from a drug discovery program for dilated cardiomyopathy using high-throughput imaging of sarcomere structure in stem cell-derived cardiomyocytes. We were able to build disease models with high accuracy, which were then deployed to identify small molecules that showed to reverse the disease phenotype. The identified small molecules were further validated with functional assays and preclinical mouse studies. Second, we present data from a pilot toxicity testing study using stem cell-derived cardiomyocytes. Our novel image-based AI method was successful in capturing dose-dependent structural changes on a panel of drugs with known cardiotoxicity profiles, while no change was detected for the negative control. The detected structural changes correlated strongly with contractility. Finally, we present data from a pilot quality control study using current-trace signals from a patch-clamp instrument. We successfully built an AI model that can accurately classify signals as good versus poor quality, which enables automated and consistent filtering of data during high-throughput experiments.
Optical Pooled Screens in Human Cells
Avtar Singh, The Broad Institute
Pooled genetic screens have been critical for the systematic identification of genes underlying cellular processes, but have thus far been limited to phenotypes defined by cellular enrichment or comparatively low-throughput single-cell molecular profiling. We have developed a method to make pooled libraries compatible with the rich set of spatially and temporally resolved phenotypes accessible to high-content microscopy by using targeted in situ sequencing to demultiplex genetic perturbations. We applied this technology to screen 952 genes for involvement in NF-κB signaling by imaging p65 nuclear translocation and relaxation, recovering most canonical pathway members and identifying novel candidate regulators of IL-1Β/TNFα-stimulated immune responses. We are currently piloting applications with a range of optical assays and cell models and expect that pooled optical screens will have broad utility in identifying genetic components, analyzing genetic circuits, and interrogating disease variants.
High-Throughput Single-Cell Imaging and Advanced MachineLlearning Supported Image Analysis of Primary Tumors Enables Anti-Cancer Therapy Development
Gregory Vladimer, Allcyte
The ability to perform high-content screening in a high-throughput fashion is routinely limited to cell lines and other explant model systems, however, there is a risk that these may not be fully representative of the in vivo environment due to culture adaptation or the lack of multi-lineage cell types. The ability to gather high-content data directly from primary samples, however, both direct from blood and bone marrow, metastasized cancers and dissociated solid tumor, without cell outgrowth or selection in a method amenable to laboratory automation can be a more direct system. Further, by combining imaging of these primary sample with an adaptable analysis pipelines robust to micro-aggregates, especially formed in solid tumor biopsy homogenates, vastly different cell shapes and sizes, and that can ultimately harness the features from each cell can become a powerful means to study drug response in a variety of indications using model systems directly derived from the patient. This methodology has been used to prioritize therapy for late-stage patients with hematological cancers in a basket trial (Snijder & Vladimer et al 2017, Lancet Hematology), has been integrated with genetic data to further uncover biological understanding and clinical synergy options (Schmidl & Vladimer et al 2019, Nat Chem Bio). Here, this talk will specifically focus on the details of the computational framework, including supervised and unsupervised machine learning approaches for cell identification and feature extraction, and other aspects of necessary infrastructure including cloud-deployment that is used to, in very high-throughput, quantify single-cell phenotypes form primary material from cancer patients for drug discovery. Further, the use case of understanding single-cell phenotypes after drug screening, both in single-cell suspensions and in micro-aggregate multi-cell / 3D environments, will be highlighted.
Session Chair: Roger Clark, B.Sc., Charles River Laboratories (UK)
A Systematic Medium-Scale Comparative Study of 2D Vs. 3D Models Using High-Content Imaging Approaches
Thierry Dorval, Institut De Recherches Servier
The development of new pharmaceutical drugs is an expensive and high-risk endeavor for the pharmaceutical industry. Major advances in physiologically relevant in vitro cellular assays such as three-dimensional models, induced pluripotent stem cells, organ-on-chip are expected to provide a better ability to predict therapeutic response, hence, reducing clinical attrition. Unlike high-throughput screening, high-content screening combines automated fluorescence microscopy with quantitative image analysis allowing phenotypic multiparametric readouts such as cell viability, DNA damage or mitochondria structure among many others. This approach is particularly well suited for complex or partially characterized targets. Moreover, in the oncology field, it has been shown that compound efficacy could be dramatically modulated in 3D models.
In this context, we are aiming to perform a medium scale screening campaign using a chemically diverse compounds collection on a cell line derived from non-small cell lung cancer with a specific mutation both in 2D and 3D models. Using high-content imaging approaches, a large set of parameters will be extracted, leading to a better characterization of various toxicity mechanisms of actions.
To provide robust comparable results between cellular models, a specific subset of compounds was selected and screened in dose responses during the assay development workflow. The analysis of this rich set of complex data provided an opportunity to improve the rest of the screening campaign.
Hits obtained from both screens will be classified, compared and validated in dose responses for a better understanding of the difference induced by the use of a 3D model combined with high-content imaging. Ultimately this could help assess the relevance of the 3D model in drug discovery in oncology.
A Novel Multiplexed uHTS and uHCS (MuHTCS) Platform in a 1536-Well Format for Chemical Biology Screening Using 3D Patient-Derived Cancer Organoids
Yuhong Du, Emory Chemical Biology Discovery Center
The current effort to grow human tissues as 3D “organoids” for cancer research aims to recapitulate 3D architecture of tumors in an in vitro environment for cancer biology studies and therapeutic development. However, due to various technical challenges, primary 3D organoid culture has not been widely used in a high-throughput screening (HTS) format for chemical screening. Here, we report the miniaturization and development of a multiplexed uHTS and uHCS (MuHTCS) organoid culturing platform for effective compound screening in a 1536-well format. Using pancreatic patient tumor-derived organoids as a model system, we optimized the 3D organoid culturing conditions with extracellular matrix (ECM). The growth of organoids was monitored by automated imaging. We further developed a multiplexed screening platform to simultaneously monitor the effect of compounds on the growth of organoids for ultraHTS (uHTS) and on the morphological change of organoids for ultra-high-content screening (uHCS) in a 1536-well plate. The MuHTCS assay has achieved Z’ > 0.5 and signal-to-background (S/B) > 6. A pilot screening of ~2000 FDA approved and the bioactive compound libraries have validated the assay for screening. Our data have demonstrated that it is feasible to utilize miniaturized 3D cancer organoids for large scale compound screening. The optimized MuHTC platform provides an efficient approach to accelerate 3-D organoids-enabled screening for drug discovery.
Acute Myeloid Leukemia Drug Sensitivity Testing Using Patient-Derived Cells in 1536 Format
Lynn Rasmussen, Southern Research
In practice, the choice of which drug to prescribe for an individual patient is often made without any information on how that individual will respond to a specific drug. The field of precision medicine is attempting to provide data to fill that information gap to match the patient with the most effective treatment for that individual. To try to fill that gap for AML patients we have developed a drug screening process using a panel of FDA approved drugs with patient-derived leukemia cells. Because the patient-derived cells are an extremely limited resource, a 1536-well assay format was developed to maximize the amount of data that could be generated for each patient. The process of developing this assay will be discussed, including the technical challenges, their solutions and the equipment choices used to achieve a reliable HTS format screening protocol.
Automating a CRISPR based rescue screen for Alzheimer’s phenotypes in iPSC derived neurons
Shushant Jain, Charles River
Alzheimer’s disease (AD) is a complex disorder with increasing prevalence and socio-economic burden. However, the majority of strategies aimed at identifying therapies for AD have been focused on targeting Abeta or TAU, which make up the plaques and tangles respectively commonly found in people with AD. The continued failure of the drug discovery process and the accompanying trials against these targets have necessitated more and better options for therapeutic intervention. Using multi-parametric high content phenotypic readouts with neurons derived from human differentiated iPSCs with familial AD mutations, we will perform CRISPR based rescue screen for the various phenotypes associated with the mutations, such as endolysosomal transport, synaptic dysfunction, neuronal toxicity. The multiple-phenotypic rescue approach will enable the identification of novel key pathways and/or targets that could serve as drug candidates for the treatment of AD.
Session Chair: Alex Godfrey, Ph.D., National Center for Advancing Translational Sciences (NCATS) (USA)
Modern Automated Chemical Synthesis and Purification of Small Molecules
Gerard Rosse, Arrival Discovery LLC
Today, drug discovery remains a game of big numbers and many organizations routinely investigate small molecules collection in the 200,000 to 2 million compound range. Automation of chemical synthesis has gained a renewed interest to produce novel compounds and to facilitate the challenging multidimensional problem of compound optimization. This presentation will describe the strategy and obstacles to implement a technology platform for the production of 90,000 compounds per year in 13 mg quantity and >90% purity. Custom designed robotic instruments, specialized laboratory infrastructure, workflows, logistics and data management to enable high throughput synthesis and purification will be discussed. The implementation of core supercritical fluid chromatography (SFC) technologies provided a unique opportunity to increase productivity and significantly reduce operational costs. The presentation will conclude with an overview of the integration of chemistry automation with advanced compounds management systems.
A Beginner’s Guide to the Practicalities of Automating Chemical Synthesis
Paul Harper, AstraZeneca
In late 2017 AstraZeneca undertook an internal “hack-a-thon”, bringing together a diverse skill set to investigate our ambition of fully automated chemical synthesis for drug-like compounds. Over the past 2 years, we have evolved 4 prototypes to better understand the challenges associated with all stages of a multi-step synthesis process. In this presentation, we will review our most recent evolution, realizing fully automated batch and flow chemistries to fuel automated synthesis. We will describe in detail the integration and optimization of the Zinsser SOPHAS platform for batch chemical synthesis; along with the varied Waters devices for product purification and analysis.
To achieve seamless integration, we have selected a third-party process scheduling software. Here we will cover the development of new drivers, protocols and interfaces for chemistry system control, along with the unusual demand of tracking and scheduling both single vial and associated plate-based activities (and the interplay between the 2 formats).
Looking forward to our 5th iteration, we’ll discuss our plan for pre-cursor storage, along with our concepts for how reactions could be constructed using the Zinsser REDI platform.
ynFini: An Automated Chemical Synthesis Platform
Nathan Collins, SRI Biosciences, SRI International
Exploration in organic chemistry is still inherently a manual process both in conducting reactions in the lab and reporting results in the written literature. Both are subject to the skill of the practitioner and the next chemist who attempts to reproduce their reported results. In an effort to improve the transferability and reproducibility of chemistry we have developed an automation platform named SynFini that automates the design, reaction screening and optimization (RSO), and production of target molecules. SynFini includes three core components. A computational tool, SynRoute, develops synthetic strategies for molecules of interest. Routes to target molecules are built by combining knowledge of known chemistry from reaction databases and new reactions predicted by machine learning. A high throughput RSO platform, SynJet, (max. throughput reaction / s) enabled by inkjet printing experimentally validates these strategies. Automated analysis of the RSO outputs in the predicted route allows for the preparation of a digital synthesis protocol that drives a benchtop multistep synthesizer, AutoSyn. AutoSyn is capable of performing solution-based synthesis routes at the milligram to gram scale. To assist medicinal chemists in drug discovery programs, artificial intelligence (AI) can be included to aid in the design, selection, and prioritization of compounds with desired properties, such as biological activity and ADMET properties, and can be interfaced with SynFini automated synthesis and testing for rapid turnaround. Examples of how each tool works independently and then how they fit together into a seamless automated solution for design, synthesis and testing of a variety of molecules are presented. How such automated processes are digitally captured and electronically transferrable for ultimate reproducibility are discussed.
Presentation Title TBD
James Beck, Eli Lilly
Session Chair: Helen Plant, B.Sc., AstraZeneca (UK)
Enabling Nontraditional Screeners from a Centralized uHTS Core
Mitchell Hull, Calibr at Scripps Research
Large, ultra-high-throughput screening systems can lead to an overreliance on simple assays and deny screening access to those with lower throughput needs. There is a temptation to forego slower, complex assays in favor of ones more amenable to HTS and to pursue familiar target families that “plug-in” to know platforms. HTS facilities, sometimes siloed within a department, are often unavailable even to those working in an HTS capable organization. As a result, some organizations have turned to multiple smaller systems designed for small-scale screening. However, data and compound management issues arise in this decentralized approach. Compound spotted assay plates, created by acoustic compound transfer platforms, can mitigate these issues. However, even combined, these cannot match the capabilities of a larger system when a larger campaign is needed. We have set up a hybrid uHTS/modular acoustic transfer platform that can act as one integrated system or three modular systems. Combining this with a high-value chemical library and an active pursuit of partners with high-value bioassays, we have pursued an approach to enable high-value, low-throughput assays, while maintaining uHTS capability and centralized compound and data management.
A B Cell Antibody Discovery Platform Using an ‘Islands of Automation’ Approach
Paul Anderson, Eli Lilly and Company
In recent years Lilly has implemented a Next-Generation Research (NGR) initiative to improve the value output of the R&D portfolio. One of the NGR pillars focuses on decreasing the timeline to bring medicines to patients. As part of this initiative, Lilly has invested in an ‘Islands of Automation’ approach to advance the B cell antibody discovery platform at the Lilly Biotechnology Center in San Diego to significantly increase throughput and reduce project timelines. This talk will focus on the automated systems that have been developed as part of this platform beginning with B cells sorted into microtiter plates through binding and functional assays run in dose-response plates from recombinantly expressed material. Examples of alternative instruments to traditional methods and software solutions to eliminate inefficiencies will be shared. This highly automated approach to antibody discovery has allowed us to meet aggressive timelines and dramatically increase throughput while allowing flexibility for future changes to our process.
Innovative Tube and Dispensing Technologies Enable Fully Acoustic Workflows for Drug Discovery Assays
Silvio Di Castro, Sample Management / Discovery Sciences / AstraZeneca
Innovative design and deployment of novel labware, instrumentation and software technologies have delivered an automated, fully acoustic platform and a step-change in small molecule Sample Management (SM) processes.
For many years, conventional SM workflows have included multiple sample transfers between vessels, using a hybrid of contact and non-contact dispensing, which are cumulatively wasteful. These combine to affect excessive sample consumption, necessitating chemists to synthesize superfluous quantities of the compound.
Here we show high-quality concordant datasets from the first fully acoustic workflow for physicochemical, enzymatic, cellular and in vitro ADME assays. We also show a reduction in (i) sample usage in these assays, (ii) DMSO usage throughout the process, and (iii) future synthesis requirements.
An acoustically compatible storage tube (FluidX™ AcoustiX™ Sample Tubes, Brooks Life Sciences, UK) was designed with optimum geometry for dispensing accuracy and speed, whilst maintaining a working sample volume able to sustain a 10-year screening lifetime. Co-molded capping technology for these tubes has resulted in increased durability for multiple dispense access and sample longevity, whilst a novel split barcode at the base affords a central opening for transmission of the acoustic pulse.
A tube-compatible acoustic liquid handler (Echo® 655T Liquid Handler, Beckman Coulter Life Sciences, USA) has been designed to utilize acoustically compatible storage tubes including a faster drop-transfer rate via a new transducer-focussing mechanism. A new dryer system removes moisture on the exterior of the tube, alongside local humidity control in the drop-transfer zone to maintain sample integrity.
The development of a new, fully acoustic workflow has minimized sample handling and waste, enabling miniaturization of assays and hence reducing the amounts of sample required for synthesis to support drug discovery projects. We have implemented and validated novel labware and instruments for a transformative and sustainable solution to many drug discovery issues applicable across the industry.
Presentation Title TBD
David Dambman, Biosero, Inc
Session Chair: Mindy Davis, Ph.D., National Institute of Allergy and Infectious Diseases (NIAID) (USA)
Continued Development of High-Throughput MS and Applications for Cell Analysis
Jonathan Wingfield, AstraZeneca
Over recent years, AstraZeneca has worked to develop a high-throughput mass spectrometry platform that utilizes the speed and contactless nature of acoustics as a sample introduction technology. Fully automated acoustic mist ionization mass spectrometry platforms are now routinely used to support biochemical HTS campaigns, to date over 10 million samples have been successfully screened against more than 10 different enzyme targets using this technology.
Having established a primary role for AMI-MS we are now looking to expand the application space where the technology could add value to early drug discovery. We have recently started to evaluate the impact of AMI-MS for metabolomic analysis of cell lysates, primarily within the early toxicology screening area.
In December 2018, AstraZeneca and collaborators at several Swedish academic institutions and SME’s were awarded a phase 2 grant from Sweden’s Innovation Agency, Vinnova. The collaboration aims to develop technologies and workflows to enable primary patient-derived disease cells to be utilized in the early phase of drug discovery.
This presentation will focus on the continued development of AMI-MS within AZ and how we are applying high-throughput mass spectrometry to enable clinical samples to be assessed in the early phases of drug discovery.
Array and Microfluidic-Based Cellular Assays Miniaturized Beyond the 1536 Well Plate
Small molecule high throughput screening (HTS) in drug discovery traditionally involves microtiter plate screening in 384- and 1536- well formats. While these methods are miniaturized compared to Petri dishes or flasks, reagent and labor costs are still significant factors in high throughput screening campaigns. Here we present the development of cellular assays utilizing ultra miniaturized array-based and microfluidic devices. The experiments were aimed at reducing cellular assay volumes from uL to nL volumes. Challenges included maintaining environmental controls for cell health and handling small volumes for cells and compounds. Novel approaches in equipment, device design, and automation were required. Data suggest that cell health, cell morphology, and pharmacological responses to drugs were similar in nL volumes compared to those observed in 50 uL volumes in 384 well plates. The development of processes and automation to industrialize new devices will ultimately enable these technologies to be applied broadly in drug discovery. Cost savings in cell and reagent usage and the ability to use disease-relevant cell systems are paths toward reduced attrition in drug discovery.
Approaches that Enable Large-Scale Chemical Biology Interrogations
Fred King, GNF Systems
During SLAS (Lab Automation) 2009 we presented a low cost, automation-friendly, screening platform that used reporter gene assays (RGA’s) to delineate the interaction of small molecules with canonical mammalian signal transduction pathways. Our experience with this approach over the last decade has demonstrated its broad utility in the support of phenotypic screening, ranging from generating mechanism of action hypotheses for individual compounds to characterizing compound libraries. Furthermore, this RGA panel inspired and guided the development of additional platforms that are more comprehensive and flexible in terms of both cell types under investigation and the scope of biological activity detected. This presentation will focus on several of these new technologies, which all leverage Next Generation Sequencing technologies to measure RNA expression levels in a multiplexed fashion. The suite of approaches provides users with the ability to balance sequencing depth, transcriptome coverage and cost per well in their assay design. Coupled with internally designed automation platforms these systems allow expression levels of thousands of genes to be monitored in every well of an HTS-sized screen.
Identification of Chemical Compounds Inhibiting Zika Virus Replication Through a Large-Scale High-Content Screening Approach
Laura Riva, Sanford Burnham Prebys Medical Discovery
Zika virus (ZIKV) is a human mosquito-borne positive-sense RNA virus, belonging to the Flaviviridae family. World Health Organization (WHO) classified this virus as an Emergency in 2016 and currently identifies Zika as a priority disease. Although symptoms are generally mild, a risk of neurologic complications including Guillain Barré Syndrome is associated with the infection in adults, while infection during pregnancy is responsible for microcephaly and other congenital malformations. Since no vaccine or commercialized antiviral targeting this virus are available, scientific efforts are currently focusing on the development of treatments allowing to efficiently limit ZIKV spread. Prompted by this unmet medical need, we conducted a screen of 51,520 small chemical compounds using a high-content imaging cell-based assay, monitoring Zika virus replication within Huh-7.5 cells by combining DAPI staining of cellular nuclei together with immunostaining of the Zika virus envelope protein. 99 candidates were identified and validated as inhibiting ZIKV replication of at least 50% at a concentration of 10 µM. Subsequent dose-response studies were performed to evaluate the effects of each compound on both virus replication and cytotoxicity and compounds showing a strong dose-response inhibitory effect on replication with weak cell toxicity were then selected for follow-up studies. Two compounds sharing a common structure presented a particularly promising antiviral activity with a selectivity index, calculated as the ratio of 50 % inhibitory (IC50) and 50 % viability (CC50) concentrations, greater than 30. This common chemical scaffold showed to specifically inhibit ZIKV, displaying an antiviral activity against several strains of both African and Asian lineages but no effect on other Flaviviruses tested. Its antiviral activity was confirmed with similar efficacy in more relevant models for ZIKV infection, including human monocyte-derived dendritic cells (hMDDCs), human neural progenitor cells (hNPC) and the placenta-derived choriocarcinoma cell line JEG-3. Time of addition kinetics as well as specific entry and replication assays excluded an inhibitory role during ZIKV entry, highlighting an antiviral role during the RNA replication step. This observation, in addition to the appearance of resistant mutant viruses upon selection in the presence of the drug, strongly suggested a non-structural protein of the virus as a target of the compound. Current efforts are ongoing to identify the specific viral target of the compound and to get more insights about its mechanism of action. In addition, Pharmacokinetics (PK) and in vivo efficacy studies will be performed in the near future to evaluate the therapeutic potential of this compound. In summary, taking advantage of a cell-based large-scale high-content screening approach to identify small chemical compounds showing antiviral activity against ZIKV, we identified a chemical scaffold specifically targeting this Flavivirus, inhibiting its RNA replication step.
Session Chair: Sam Michael, National Center for Advancing Translational Sciences (NCATS) (USA)
Improving Daily Operation of a Fully-Automated uHTS System
Steven van Helden, Pivot Park Screening Centre
Pivot Park Screening Centre (PPSC) is a small company that specializes in ultra-high-throughput screening (uHTS) for drug discovery. We perform about 25 full deck screening campaigns a year on our own library of 300.000 compounds, the European Lead Factory library of 550.000 compounds and client libraries up to 1.000.000 compounds. In order to support this huge production, we have implemented efficient processes on a fully automated screening system consisting of 3 robot pods integrated with a wide variety of instruments.
Even in a highly automated environment like our uHTS lab, there is a continuous need for tools and little tricks to support our daily work. These include software tools to run active picking from the online store, re-use of (washed) verification plates, a simple tool for drying compound plates, implementation of a weighing station in the robot for checking dispensing performance, etc. Also, we have implemented cleaning stations for our heavily used certus dispensers. Finally, we make extensive use of a 3D-printer to create all sorts of tools in the lab and to save costs by printing parts of instruments that need repair. This presentation will provide insight into the daily operations in a uHTS lab.
Leveraging Open Source Electronics for Rapid Development of Custom Laboratory Devices
Pierre Baillargeon, The Scripps Research Institute
The Lead Identification team at Scripps Florida routinely leverages open source technologies to meet operational challenges and to provide custom engineering tools for lab use. Most recently, these tools include the development of a micro solenoid dispensing QC platform built around the Lee Company VHS series valve. This system was developed to address the unmet need of dispensing of 3D models such as spheroids and to assist with general QC of valves used in ongoing HTS efforts. This QC platform allows users to characterize & optimize the performance of VHS valves under a variety of conditions.
This reconfigurable QC platform is built on an optical breadboard and is comprised of three main subsystems: electronics, optical train and motion control. The electronics subsystem allows users to easily control VHS series valve behavior using an Arduino microcontroller through a custom in-house designed Arduino shield. The optical train consists of an off the shelf USB camera combined with an in-house designed open-source illumination panel that allows imaging of individual droplets via the stroboscopic effect. An open-source X/Y motion control system further increases the utility of the platform by allowing automated dispensing into microplates. User control of the QC platform is provided via a custom web-based interface that communicates directly with the microcontroller and allows users to easily specify microplate dispense patterns by interacting with graphic representations of microplate wells.
Also presented is the development of a custom Arduino-based syringe pump system intended for use alongside the Lee valve QC platform. This syringe pump system utilizes a Tecan Cavro pump and allows for real-time adjustment of pump parameters during mixing and dispensing operations which are critical for mixing and homogeneous distribution of the spheroids. The development of these platforms, lessons learned and results of initial testing are presented.
Intelligent Microscopes Using Open-Source Hardware for High-Throughput Laboratory Automation
Pavan Chandra Konda, Duke University
Traditional microscopes used for automated imaging and analysis sets one aback with tens of thousands if not hundreds of thousands of dollars. This limits the number of microscopes a lab can afford, hence limiting the number of parallel experiments that can be performed. We present a novel approach by combining low-cost, low-resolution microscopes with advanced computational imaging methods that can extract high-resolution image information in the post-processing. In addition, we implement novel machine learning methods to jointly optimize the automation task, e.g. cell segmentation, and the data acquisition process, e.g. illumination pattern, to capture fewer data without losing the performance of the automated task.
Our initial prototype costing ~$150 employed a Raspberry Pi as the computer and a modified Raspberry Pi V2 camera as the low-resolution microscope. A low-cost 16x16 LED array developed for display is used to illuminate the sample and 3D printed parts are used for assembly. LEDs in the array are sequentially illuminated to capture 256 low-resolution images, where the high-resolution information is encoded within these low-resolution images using the aperture synthesis concepts. The captured 256 low-resolution images were combined to achieve 0.8µm resolution, for the first time in a low-cost setting, across 4 mm2 field-of-view. The phase of the object is also recovered in the process, making this suitable for imaging cell cultures without any need of staining. In the latest developments, we implemented a new machine learning model to multiplex the illumination to reduce the number of images captured to two, without any loss in performance for tasks such as cell segmentation or detecting malaria infection. This also reduces the image processing time and, exploiting the increasing computing performance on opensource hardware such as Raspberry Pi and Google’s Coral Edge TPU, we are currently working towards achieving real-time machine learning-based automation on our portable low-cost setup.
The 3D printed design of our microscope can be easily modified to the specific requirements of a lab, e.g. imaging stress fiber reorientation in cells under mechanical stimuli requires a different setup compared to imaging cell confluency in a petri-dish. Our optics and algorithms still stay valid for all these different configurations and the required modifications in the 3D printed designs are usually minor. This is not possible with commercial systems that are designed for a limited number of imaging applications. Combining the latest developments in machine learning makes our approach a powerful tool for laboratory automation and diagnostics in low-resource settings.
Beyond High-Content Screening: An Open Next Generation Image Analysis Platform
Peter Bajcsy, NIST National Institute of Standards and Technology
There is an increasing interest in discoveries from images acquired by high-throughput and high content microscopy imaging of multi-well plates with biological specimens under a variety of conditions. As multi-dimensional automated imaging increases its throughput to thousands of images per hour, the computational infrastructure for handling the images has become a major bottleneck. The bottleneck associated challenges arise due to big image data, complex phenomena to model, non-trivial computational scalability that leverages the advanced hardware and cutting-edge algorithms, and incompatible software tools that vary in the language they were written in, the platform they were written for and capabilities they were designed to execute.
To address the above challenges, groups have developed software solutions based on client-server systems with modern web technologies on the client-side and a spectrum of databases, computational workflow engines, and communication protocols on the server-side to hide the infrastructure complexity. However, these solutions have not focused on the interoperability of imaging specific computational plugins and visual exploratory capabilities of such plugins over very large image collections.
To address these inter-operability and visual exploration challenges, the National Institute of Standards and Technology (NIST) and the National Institutes of Health (NIH) - National Center for Advancing Translational Science (NCATS) have formed a close collaboration to develop an open-source platform for executing web-based image processing pipelines over very large image collections with interoperable plugins. The plugins developed by both institutes are based on software containers as standardized units for server-side deployment, as well as on dynamically created web user interfaces (UI) to enter parameters needed for the software execution and for advanced visual data explorations on the client-side. Each container packages code, with all its dependencies, and has an entry point for running the computation in any computing environment. Each UI description file contains metadata about the plugin-container and the computation parameters.
We will demonstrate the utility of the platform with algorithmic plugins by analyzing 1536 well plates with three spectral channels and multiple fields of views (FOVs) per well for drug dose-response across an array of features. Typical visual data exploration is assisted by algorithmic tools for quality control, stitching of FOVs per well, segmentation, characterization of regions of interest, and scalable visualization using Deep Zoom, a toolkit for browser viewing of gigapixel 2D images. The data explorations are interactive either in a Deep Zoom viewer or in a Jupyter notebook while prototyping pipelines. More demanding computations are supported via batch processing and deep learning-based pipelines are designed for GPU execution. With the NIST and NIH NCATS combined efforts, researchers are enabled to discover quantitative insights from their imaging data and reuse computational tools developed by anyone following the web computational plugin conventions.
Track Chairs: Rob Howes, Ph.D., AstraZeneca (UK) and Janice Reichert, Ph.D., The Antibody Society (USA)
Session Chair: Janice Reichert, Ph.D., The Antibody Society/mAbs (USA)
Creating a Novel T-cell Engaging Bispecific Antibody Platform: Fine-Tuning Anti-Tumor Activity with Sequence-Based Discovery and Machine Learning
Katherine Harris, Teneobio
Using a unique sequence-based discovery approach along with proprietary transgenic rats, we have created a large collection of fully human anti-CD3 antibodies with diverse T-cell agonist activities. Our novel discovery platform combines antibody repertoire deep sequencing, high-throughput gene assembly, and recombinant expression. Our approach generates a large diversity of sequence-defined antibodies that we characterized in high-throughput T-cell agonist assays. Using machine learning tools, we were able to rapidly establish sequence-activity relationships and identify key residues and variable region positions in the antibody repertoire that had desired agonist behavior. The CD3 antibodies identified by our platform show diverse in vitro T-cell activation profiles measured by CD69 upregulation, IL2, and IFNg production. We also generated human domain antibodies targeting a variety of tumor antigens that we combined with our unique CD3 antibodies to create bispecific molecules that mediate redirected T-cell killing of tumor cells. In one particular example, we have created a panel of aCD3:aBCMA bispecific antibodies for the treatment of multiple myeloma that stimulate different levels of T-cell activity. Using a multiple myeloma tumor cell line along with primary human PBMCs, we demonstrate a spectrum of in vitro tumor cell killing activity with varying levels of cytokine release using our bispecific molecules with diverse CD3 binding activities. In summary, we have created a T-cell engaging bispecific antibody platform with tuned T-cell agonism that can be used to optimize the therapeutic index for a variety of tumor antigens.
Benefits of Chicken-Derived Antibodies for Combination Immunotherapy
Torben Gjetting, Symphogen
The development of novel antibodies and more powerful therapeutic combinations for immunotherapy is an intense area of focus. However, for difficult and/or conserved targets, finding antibodies with unique functionality, and generating early proof of concept pose challenges to the development of novel antibody therapeutics. Symphogen’s approach to discovery and development of potent antibody combinations for cancer immunotherapy using different species, including chicken, will be presented. Examples from our clinical pipeline will be shown.
Quantitative High-Throughput Screening Assays for the Discovery and Development of SIRPα-CD47 Interaction Inhibitors
Thomas Miller, Institut Paoli Calmettes
CD47 is an immune checkpoint molecule that downregulates key aspects of both the innate and adaptive anti-tumor immune response via its counter receptor SIRPα, and it is expressed at high levels in a wide variety of tumor types. This has led to the development of biologics that inhibit SIRPα engagement including humanized CD47 antibodies and a soluble SIRPα decoy receptor that are currently undergoing clinical trials. Unfortunately, toxicological issues, including anemia related to on-target mechanisms, are barriers to their clinical advancement. Another potential issue with large biologics that bind CD47 is the perturbation of CD47 signaling through its high-affinity interaction with the matricellular protein thrombospondin-1 (TSP-1). One approach to avoid these shortcomings is to identify and develop small molecule molecular probes and pretherapeutic agents that would (1) selectively target SIRPα or TSP-1 interactions with CD47, (2) provide a route to optimize pharmacokinetics, reduce on-target toxicity and maximize tissue penetration, and (3) provide for more flexible routes of administration. As the first step toward this goal, we report the development of an automated quantitative high throughput screening (qHTS) assay platform capable of screening large diverse drug-like chemical libraries to discover novel small molecules that inhibit CD47-SIRPα interaction. Using time-resolved fluorescent resonance energy transfer (TR-FRET) and bead-based luminescent oxygen channeling assay formats (AlphaScreen), we developed biochemical assays, optimized their performance, and individually tested them in small-molecule library screening. Based on performance and low false-positive rate, the LANCE TR-FRET assay was employed in a ~90,000 compound library qHTS, while the AlphaScreen oxygen channeling assay served as a cross-validation orthogonal assay for follow-up characterization. With this multi-assay strategy, we successfully eliminated compounds that interfered with the assays and identified five compounds that inhibit the CD47-SIRPα interaction; these compounds will be further characterized and later disclosed. Importantly, our results validate the large library qHTS for antagonists of CD47-SIRPα interaction and suggest broad applicability of this approach to screen chemical libraries for other protein-protein interaction modulators.
Discovery of High-Affinity, Pan-Allelic, and Pan-Mammalian Reactive Antibodies Against the Myeloid Checkpoint Receptor SIRPα
Janet Sim, ALX Oncology
Targeting the CD47-signal-regulatory protein α (SIRPα) pathway represents a novel therapeutic approach to enhance anti-cancer immunity by promoting both innate and adaptive immune responses. Ongoing clinical trials to inhibit this pathway by targeting CD47 has shown promising results in reducing tumor burden (Chow et al., ASCO 2019). Unlike CD47 which is expressed ubiquitously, SIRPα expression is mainly restricted to myeloid cells and neurons. Therefore, compared to CD47-targeted therapies, targeting SIRPα may result in differential safety and efficacy profiles, potentially enabling lower effective doses and improved pharmacokinetics and pharmacodynamics. In this talk, we will present our strategies of using wildtype/human antibody transgenic chickens for immunization and describe our screening approaches to identify SIRPα antibodies suitable for clinical translation. A total of 200 antibodies were isolated and approximately 70 antibodies with diverse SIRPα binding profiles, sequence families, and epitopes were characterized. A subset of antibodies was shown to bind both human SIRPα v1 and v2 alleles with high affinity (nM-pM), potently antagonize the CD47/SIRPα interaction and potentiate antibody-dependent cellular phagocytosis in vitro. The anti-SIRPα antibodies also enhanced anti-tumor activity in both xenograft and syngeneic tumor models and were well tolerated in cynomolgus monkeys with favorable PK and extended receptor occupancy. These properties provide an attractive rationale to advance the development of these anti-SIRPα antibodies as a novel therapy for advanced malignancies.
Session Chair: David Gilham, Ph.D., Celyad (BEL)
Generation of a Controllable CAR T Cell Therapy
Travis Young, Calibr, a Division of Scripps Research
T cell-based therapies, including T cell engaging bispecific antibodies, and genetically engineered chimeric antigen receptor engineered T cells (CAR-T cells) have produced remarkable results in clinical trials – achieving complete remissions in patients with hematological malignancies who failed multiple lines of prior therapy. Towards increasing the potency and safety of these therapeutics, as well as expanding them to cancers outside of the hematological space, we have defined how the biophysical characteristics of the antibody-based components of these therapies modulate the physiological response of the T cells that carry out the anti-tumor activity. For example, we have designed a “switchable” CAR-T cell system using antibody-based molecular switches. This platform enables fully tunable control of CAR-T cell activity in a universal format that can be redirected to nearly any therapeutic antigen target. The platform is expected to reduce the risk of severe adverse events that have plagued the development of CAR-T cell therapies clinically. We have demonstrated such a platform can reduce risks related to cytokine release syndrome and double as a safety switch to turn the therapy off in the case of an adverse event. We have further demonstrated the temporal control over CAR-T cell activation enables the development of robust central memory T cells. In preclinical mouse models, we’ve demonstrated these central memory cells can be recalled affording on-demand, in vivo T cell expansion. These concepts are expected to be central to clinical efficacy with T cell-based therapeutics and important to ultimately achieving efficacy in solid tumors. A proof of concept clinical trial will be initiated in late 2019 for patients with lymphoma.
Pluripotent Cell-Derived Engineered T and NK Cells as a Cornerstone Approach for Off-the-Shelf Cancer Immunotherapy
Bob Valamehr, Fate Therapeutics
Several obstacles currently hamper the broad use of adoptive cell therapies, including the inherent variability and cost of manufacturing of cellular populations, the absolute requirement for precise genetic editing of multiple elements and the elimination of undesired stochastic events associated with cellular engineering. Here we present a unique approach to create master pluripotent stem cell lines, clonally derived to contain precisely edited events at the single-cell level and the conversion of that master cell line into uniform populations of highly efficacious off-the-shelf engineered T and NK cells.
Exploiting Natural Killer Receptors for Autologous and Allogeneic CAR T Cell Therapy of Cancer
David Gilham, Celyad
Chimeric Antigen Receptor (CAR) T-cell therapy has hit the headlines with impressive clinical responses in hematologic B-cell malignancies that have led to the successful licensing of two products that both target CD19. Anti-BCMA CAR T-cell therapies for myeloma might come next but there is a dearth of targets outside of the B-cell malignancy space.
Celyad has been exploring the potential of Natural Killer cell receptors to target cancer. Specifically, the company is conducting a series of clinical trials testing the safety and efficacy of CAR-T cells bearing the Natural Killer Group 2D (NKG2D) receptor that can specifically bind eight stressed induced ligands found on a broad range of cancers, yet largely absent from the surface of non-malignant, healthy cells.
The first trial involved giving multiple infusions of autologous NKG2D-based CAR-T cells without pre-conditioning chemotherapy and provided some initial evidence of clinical activity with a good safety profile. Further clinical activity has tested this CAR-T approach with pre-conditioning and standard of care chemotherapy in hematological and solid tumors. Results of these trials and the company’s trial of allogeneic NKG2D-based CAR T cell therapy will be reviewed; the latter being thought to be the first allogeneic CAR-T approach to be tested in the solid cancers.
Aside from NKG2D focused studies, the company is also embarking on a broader allogeneic CAR T cell platform technology exploiting interfering RNA to control graft versus host disease, the primary limitation of using allogeneic cells for therapy.
Taken together, these early clinical studies suggest that the NKG2D receptor in both autologous and allogeneic approaches could provide the potential to target a broad range of tumor-targeting that could follow the success of CD19 and BCMA CAR T cell therapy. The challenges ahead relate to how to exploit the targeting ability of NKG2D. Some of the next steps including manipulating the memory phenotype of the CAR T cell product will be discussed.
Dynamic Profiling of Antitumor Activity of CAR T Cells Using Micropatterned Tumor Arrays
Xiao Wang, Harvard Medical School
Cancer immunotherapy based on Chimeric antigen receptor (CARs) T cells has emerged as one of the most promising new therapies for the treatment of patients with B-cell malignancies. To address the critical need for preclinical assessments of essential CAR-T cell functions, we designed an assay for the dynamic and high-content profiling of trafficking and antitumor activity of CAR-T cells. Specifically, we micropatterned multiple myeloma tumor cells into arrays of microscale islands and housed them with CAR T cells in microfluidic chambers. The interaction of CAR T cells with the tumor islands was captured with time-lapse microscopy which generated multifaceted information about CAR T trafficking and tumor elimination over time. Spatially patterning the tumor cells into hundreds of microscale islands allows for robust measurements in highly reproducible conditions. Our assay overcomes the limitations of widely-used chromium-release and luciferase-based assays which can only quantify bulk cytolytic responses at a single time point. Using the assay, we found that CAR T cells robustly migrate towards the tumor-cell islands, aggregate into large clusters that envelop the tumor cells, increase the local effector-cell density, and exert cytolytic effects on the tumor cells. We discovered the differences in trafficking, clustering, and cytotoxicity of CAR T cells from different donors. Using the assay, we also conducted multi-faceted characterizations of CAR T cells with different constructs. We compared anti-BCMA and APRIL-based CAR T cell constructs against BCMA+ and BCMA knockout (KO) multiple myeloma MM.1s tumor cells. We found that APRIL-based CAR T cells efficiently clustered around and eliminated both tumor cell types, suggesting that this CAR T cell construct could reduce the incidence of antigen-negative escape. With the ease of use, the scalability and the reproducibility and the ability to dynamically map the antitumor activity of CAR T cells, the micropatterned tumor array (MiTA) could become a useful tool for studying cancer immunology and aiding the pre-clinical evaluations of cell-based cancer immunotherapy.
Session Chair: Rob Howes, Ph.D., AstraZeneca (UK)
Target and Drug Discovery in ‘Undruggable Space’ Using Functional Proteomics
Markus Muellner, PhoreMost
Despite an increasing spend on drug development, many diseases remain unaddressed with little hope of finding new treatment options by conventional means. One reason for this is the lack of knowledge regarding which proteins are druggable, and which pockets on the protein surface might be most beneficially targeted by small molecules. Current methods for discovering new drug targets rely on genetic knock-out (CRISPR) or knock-down (RNAi) methods. While these techniques can be useful in providing candidate therapeutic target genes, the next step of developing protein-targeting therapies often stalls due to insufficient information on druggability. To address this problem, PhoreMost has developed a functional proteomics / phenotypic screening technology called “Protein Interference” (PROTEINi®) that yields both the target’s identity as well as information on available druggable sites within the target. PROTEINi utilizes proprietary large ( >1 Million), diverse, lenti-encoded libraries of small, self-folding, three-dimensional peptide "shapes", which are expressed in live cells and, much like small molecules, interfere and engage with available pockets on target proteins on a proteome-wide scale. In contrast to shRNA or CRISPR screens, PROTEINi works on the same level as most small molecules (the proteome) and is not influenced by gene copy number, SNPs or genetic buffering. The process discovers novel targets for a given assay system as well as peptides engaging this target functionally as a starting point for drug discovery. PhoreMost currently has internal small molecule programs in Oncology, Immuno-Onc and Neurodegenerative disorders. We have recently also expanded the method into Targeted Protein Degradation space to systematically discover novel and functional E3 linkage sites across a set of 600 ligases.
Presentation Title TBD
Rob Howes, AstraZeneca
µ-Hydroporator: A Next-Generation Intracellular Delivery Platform
Aram Chung, Korea University
The introduction of biomolecules and functional nanomaterials into cells is a crucial task in diverse biological situations, including immunotherapy, genome editing, regenerative medicine, and fundamental biological studies. Traditionally, intracellular delivery is achieved by carrier-based or membrane-disruption-based techniques. Carrier-based approaches utilize reconstituted viruses (e.g., lentivirus or AAV), or liposome (e.g., Lipofectamine), and when optimized they offer effective delivery (e.g., DNA delivery for cell transfection). However, carrier-based approaches critically suffer from toxicity, low-throughput, and require time-consuming and/or labor-intensive preparation steps. Alternatively, membrane-disruption-based methods such as electroporation and microinjection create transient discontinuities on the cell membrane for target material diffusion. The physical cell membrane disruption is relatively independent of target and cell type, but they cause excessive damage to cells and suffer from limited throughput. To address these drawbacks, recent advancements in microfluidics and nanotechnologies have provided new solutions; however, identifying an ideal method that offers easy, low-cost, highly efficient, high-throughput, noninvasive and cell type/target independent delivery, remains challenging. Here, we present a next-generation intracellular delivery platform termed “µ-Hydroporator,” which introduces macromolecules into any cell type, at high-throughput, in a single-step, without a vector or external apparatus. µ-Hydroporator is purely based on the hydrodynamic cell deformation-restoration process, which opens the cell membrane and enables efficient transport of external target biomolecules or functional nanomaterials into the cell. In brief, the cell suspension mixed with target materials is injected into a T-junction microchannel with a micro-cavity where inertial vortices instantaneously deform cell. This rapid hydrodynamic cell deformation creates transient nanopores on the cell membrane, allowing the convective transport of foreign target molecules during the cell restoration process. Using µ-Hydroporator, we have successfully delivered diverse macromolecules (e.g., RNAs, Plasmids, DNAs, DNA origami, CRISPR-Cas9s, proteins, Q-dots, AuNPs, etc.) into various cell lines including difficult-to-transfect primary cell lines such as stem and immune cells, achieving highly efficient intracellular delivery (< 98%) in a high-throughput manner (~1,600,000 cells/min) while maintaining high cell viability (< 95%). Unlike traditional methods that rely on external apparatus, and/or chemical modification of target molecules, µ-Hydroporator only requires a syringe pump (not even a microscope!). This permits easy, robust and simple operation and cost-reduction from not requiring a skilled technician and instrument. We firmly believe that the reported µ-Hydroporator will establish a new paradigm in intracellular delivery, which will immensely benefit cellular engineering research and industry.
High-Throughput Encapsulation and Selection of Cells Based on Antibody Secretion Using Lab-on-a-Particle Technology
Joseph de Rutte, University of California, Los Angeles
We introduce a new approach to collect and quantify single-cell secretions without crosstalk in monodisperse droplets formed by precisely structured microparticles, enabling high-throughput screening based on this critical cell function. The ability to analyze and sort cells based on secretions (antibodies, cytokines, proteases, or other enzymes) has implications in understanding cellular heterogeneity fundamental to biology and creating new biotechnology products, such as biologics and cell therapies. Recently, droplet microfluidics has emerged as a powerful approach to perform single-cell secretion screening in high-throughput, using compartmentalization in a small volume to accumulate secreted factors to high levels for accurate detection. Despite this utility, the necessity of specialized equipment and expertise on the end-user hinders its widespread adoption. A platform that is fully compatible with standard lab equipment (e.g. pipettes, flow cytometers) has the potential to dramatically extend the reach of single-cell screening technology. Our particle-templated droplet, i.e. “Dropicle”, approach is unique in that pre-fabricated particles are used to form monodisperse emulsions that encapsulate single cells, requiring only standard lab equipment for the end-user. Cavity-containing microparticles are loaded into well plates and due to their morphology settle upright with their cavities exposed. Cells are loaded into the microparticle cavities and adhere via integrin binding sites. Biocompatible oil and surfactant are added and the suspension is agitated by pipetting to create incrementally smaller water-in-oil droplets. These resulting dropicles are monodisperse, maintaining a size defined by the particle geometry (CV < 6%), while the excess fluid is partitioned into surrounding smaller satellite droplets. Secretions from encapsulated cells are captured on the associated particles via protein A binding sites. Particles and associated cells and secretions are transferred back to the aqueous phase enabling downstream labeling and screening with standard flow cytometers. It was observed that seeded cells filled the cavities of the particles according to single-poisson statistics (in contrast to typical double-poisson statistics for single-cell, single-particle pairs in drops). After dropicle formation cells maintained high viability over 24 hours ( >80%). Initial tests with anti-IL-8 producing CHO cells demonstrate the ability to capture and label secretions on particles containing cells without crosstalk to neighboring particles. Further, we demonstrate the ability to isolate cells associated with high anti-IL-8 signals in high-throughput using commercial flow cytometry systems ( >100 sorts/s). Using this dropicle platform, researchers can perform droplet-based assays using standard lab equipment without sacrificing the precision of droplet microfluidics. Since dropicles are formed simultaneously, compartmentalization is rapid ( >400k in 30s) and can be easily scaled to accommodate large population screens. Further, the associated particle enables additional functionality such as physicochemical cues or cell-specific capture antibodies to select out specific cell types. Our results demonstrate new capabilities for lab-on-a-particle technologies that can accelerate the automation of single-cell assays.
Track Chairs: Jason Ekert, Ph.D., MBA, GlaxoSmithKline (USA) and Nancy Allbritton, M.D., Ph.D., University of North Carolina Chapel Hill (USA)
Session Chair: Nancy Allbritton, M.D., Ph.D., University of North Carolina Chapel Hill (USA)
Advanced Bioprinting Strategies for Tissue and Tissue Model Fabrication
Y. Shrike Zhang, Harvard Medical School
Bioprinting has recently emerged as an enabling technology in tissue biofabrication at high fidelity. This talk will discuss our recent efforts on developing a series of advanced bioprinting strategies, including sacrificial bioprinting that allows generation of perfusable microchannels embedded in hydrogel microchannels, microfluidic and hollow fiber bioprinting that achieves production of standalone cannular tissues, and multi-material bioprinting based on both extrusion and stereolithographic modalities that enables creation of complex hierarchical tissue microstructures. Innovations in various cytocompatible and bioactive bioink formulations will also be presented. These platform bioprinting methods have been demonstrated to facilitate faithful fabrication of biomimetic tissues and their models spanning from the heart, liver, and musculoskeletal system to blood vessels and beyond, as well as their diseased forms, for applications in regenerative medicine and inaccurate screening of therapeutic agents.
Wound-Conformal Delivery of Dermal Tissue Constructs for Full-Thickness Burn Treatment Using a Handheld Bioprinter
Richard Cheng, University of Toronto
Full-thickness burns where both the dermal and epidermal layers of the skin are destroyed result in high patient mortality due to infection, dehydration, and shock. The current standard of care involves the direct application of an acellular crosslinked protein scaffold which forms a temporary physical barrier and promotes host cell migration into the wound area; however, this is problematic in severe burns where little healthy skin is available for repair. Delivery of patient-derived autologous or immunoprivileged allogeneic cells are emerging as potential treatment options due to continuous extracellular matrix remodeling and persistent cell signaling, but challenges include homogenous delivery of cells onto a large, non-flat wound topography. Although approaches such as cell spraying and microparticle injecting have been explored in the field, the continuous formation of three-dimensional, hydrogel-based tissue constructs uniformly on a physiological wound surface remains unsolved. Here, we report the development of a handheld bioprinter that delivers wound-conformal dermal tissue constructs to improve wound healing in full-thickness burns. Mesenchymal stromal cell (MSC)-containing fibrinogen bioink and thrombin crosslinker solutions were delivered through on-board syringe pumps to a microfluidic printhead with internal bifurcated channels. Dermal tissue constructs of consistent thickness covered with the crosslinker were obtained at the exit. Wound-conformal delivery of these MSC-laden dermal tissue constructs was achieved by translating the printhead along the wound surface by a soft silicone wheel, while a two-axis gimbal design allowed it to adapt to the wound topology. We observed that the addition of 1% hyaluronic acid (HA) provided desirable shear-thinning behavior of the bioink (1.2 Pa•s at shear rate 1/s; 0.35 Pa•s at shear rate 100/s), resulting in 83% of the starting thickness to be maintained for deposition surfaces with inclination angles of 45 degrees. Furthermore, these fibrin-HA hydrogels maintained high biocompatibility with the co-delivered MSCs ( >94%), in addition to the long-term preservation of 3D morphology and cell proliferation as shown with Hoechst/Phalloidin+ immunostaining over one week. To demonstrate the clinical utility of this approach, we uniformly distributed 1x10^6 MSCs/ml of the fibrin-HA hydrogel on a porcine 5cm x 5cm full-thickness burn wound model and quantified a 1.4-fold improvement of macroscopic re-epithelialization speed, a 1.3-fold increase in collagen density in the dermal layer, and a 2.5-fold reduction in CD11b+ inflammatory cell activity after 28 days compared to burn controls, as observed via microscopic analysis of H&E histological stains. Taken together, we have shown that the handheld bioprinter can conformally deliver MSC-containing dermal tissue constructs directly on wound substrates with physiological topographies, leading to full-thickness burn wound repair as shown in porcine pre-clinical case studies.
Generation, Validation and Application of Induced Pluripotent Stem Cell Models for Functional Genomics
Lisa Mohamet, GSK
Myeloid cells play critical roles in adaptive and innate immunity and dysregulation can result in disease pathology, such as neurodegeneration. However, a detailed mechanistic understanding of human myeloid biology has been hampered by the lack of robust and scalable models for cellular and genetic studies. Conventional approaches rely upon immortalized cells that lack biological relevance or primary cells which are limited in number, reproducibility, and genetic perturbation. To overcome these challenges, we developed and industrialized a human induced pluripotent stem cell (iPSC)-derived myeloid platform that permits a robust and continuous supply of progenitors that are subsequently differentiated into macrophage or microglia. Since each iPSC line retains the genetic information of the donor this provides an opportunity to harness human genetics to investigate in vitro disease mechanisms.
We performed extensive transcriptomic, epigenetic, proteomic and metabolomic analyses with concomitant phenotypic (e.g. flow cytometry, image analysis) and functional assays (e.g. phagocytosis, cytokine secretion) to support their use as a model to primary counterparts. Here, we demonstrate a combination of conventional and innovative technologies to generate and validate iPSC-derived target cell types as an unlimited source of patient genotype-specific cells to study. We describe the implementation of such disease-relevant models to enable large scale (epi)genomic functional modeling for improved novel target ID.
The human biological samples were sourced ethically and their research use was in accord with the terms of the informed consents under an IRB/EC approved protocol.
High-Throughput Organoid and Monolayer Platforms to Study Intestinal Physiology
Scott Magness, UNC-Chapel Hill
Intestinal organoid technologies have revolutionized culture models to study physiology, disease, and injury in vitro. While primary stem cell-driven organoid cultures offer many improvements over conventional cancer cell line models, individual organoids are highly heterogeneous in lineage ratios, morphologies, growth properties, and other physiological parameters. Additionally, the enclosed lumen prohibits easy access to the apical cell surface to study nutrient absorption, the microbiome, and drug interactions with the epithelium. We have developed platforms that address these challenges. Specifically, our group focuses on engineering high-throughput systems to study single-cell stem cell biology, stem cell niche co-cultures, organoid dynamics, luminal physiology, and the microbiome. These platforms can be applied to organoids across any tissue type, are scalable, portable, and represent a high-resolution and statistically robust solution for preclinical models of human disease.
Session Chair: Deok-Ho Kim, Ph.D., Johns Hopkins University School of Medicine (USA)
Modeling Dystrophic Cardiomyopathy on a Chip for Phenotypic Drug Screening
Deok-Ho Kim, Johns Hopkins University School of Medicine
Directed differentiation of human pluripotent stem cells (hPSCs) into cardiomyocytes typically produces cells with structural, functional and biochemical properties that most closely resemble those present in the fetal heart. Here we establish an in vitroengineered developmental cardiac niche to produce matured hPSC-derived cardiomyocytes (hPSC-CMs) with enhanced sarcomere development, electrophysiology, contractile function, mitochondrial capacity and a more mature transcriptome. When this developmental cardiac niche was applied to dystrophin mutant hPSC-CMs, a robust disease phenotype emerged, which was not observed in non-matured diseased hPSC-CMs. Matured dystrophin mutant hPSC-CMs exhibited a greater propensity for arrhythmia as measured via beat rate variability, most likely due to higher resting cytosolic calcium content. Using a custom nanopatterned microelectrode array platform to screen functional output in hPSC-CMs exposed to our engineered developmental cardiac niche, we identified calcium channel blocker, nitrendipine, mitigated hPSC-CM arrhythmogenic behavior and correctly identified sildenafil as a false positive. Taken together, we demonstrate our developmental cardiac niche platform enables robust hPSC-CM maturation allowing for more accurate disease modeling and predictive drug screening.
Engineering the Cell Niche to Enhance iPSC-derived Cardiomyocyte Maturity and Predictivity in a High-Throughput, Assay-Agnostic Manner
Nicholas Geisse, NanoSurface Biomedical
Stem cell technology holds great promise for mitigating the cost of drug development by properly modeling human biology in vitro. Assays based on these technologies have the potential to be more predictive of toxicity and efficacy when compared to simpler single-molecule assays. Building these representative organ-on-chip models relies on generating hierarchically organized cells and tissues. In vivo, this organization is driven by a complex interplay of cells and their environment, including the extracellular matrix (ECM). Traditional cell culture environments—typically composed of hard and unstructured glass or plastic— fail to fulfill the role of the ECM in development. Consequently, many in vitro stem cell models often fall short incorrectly reproducing critical in vivo phenotypes because cultured cells oftentimes lose type-specific characteristics or express phenotypes indicative of an immature developmental stage. Considerable effort is directed at fabricating biomimetic culture environments that maintain or promote mature phenotypes. However, making biomimetic substrates typically involves costly or hard-to-reproduce techniques that are often incompatible with many standard assays; these challenges are compounded when using high-throughput techniques. Our objective is to develop novel surfaces that mimic the mechanical and structural cues of the ECM without compromising compatibility with state-of-the-art assays and instruments. The fabrication scheme is based on high-precision photolithography techniques and is thus highly reproducible, scalable and amenable to integration with most industry-standard endpoint assays, including high-NA microscopy. Further, the approach is centered on SLAS/ANSI/SBS-compliant formats that are compatible with high-throughput automated platforms. Our data demonstrate that various cell types are amenable to this approach. hiPSC-derived cardiomyocytes (CMs) showed in vivo-like myofibril alignment, sarcomere spacing and width, and expression of CM-specific proteins that are present in mature myocytes. Higher-ordered 2D anisotropic myocyte tissues also showed adult-like structure and electrophysiological responses to drugs in vitro when compared to traditional unordered 2D isotropic constructs. Examples of phenotype enhancement of other adherent mammalian cell types will be presented, further demonstrating the utility of the approach for generating more representative cells and tissues. Finally, we extend this technique to pattern the surface of microelectrode arrays and demonstrate that these ECM-based cues enhance the electrophysiological response of cardiomyocytes to various drugs of known action. We conclude that our approach is a viable method for re-creating specific aspects of the ECM that are critical for driving the development and maturation of stem cells in culture.
Automation of Multi-Organ Chip Assays and Microscopic Analysis
Ann-Kristin Muhsmann, Technische Universität Berlin
Microphysiological systems (MPS) are designed to mimic human organs and physiology with the aim of improving the drug development process. They have proven to be a powerful tool at the research level and a solid basis for the establishment of qualified preclinical assays with improved predictive power. However, one remaining drawback is the lack of comprehensive standardization. This hinders their use for regulatory purposes as well as it impedes the extraction of the full extent of information from acquired data. Moreover, one main advantage of MPS in contrast to animal testing – the insight in “the body” throughout the test assays – cannot be deployed fully as long as assay execution, observation and analysis are highly time-consuming and resource binding. This is why there is an immediate demand to automate MPS cultivation and analysis. Concluding, we developed a dedicated system for the cultivation of our TissUse Multi-Organ-Chips.
The Humimic AutoLab cultivates up to 24 Multi-Organ-Chips (MOCs) simultaneously, providing customized incubation and systemic pulsatile media circulation. Regular media exchanges, substance application and sample extraction are executed automatically according to assay specifications. Routine microscopic analyses such as bright field imaging and fluorescence measurements are conducted in scheduled cycles and at any chosen time. To hinder contamination and for operator protection, all liquid handling steps are conducted under sterile conditions. Media, substances and samples are stored in a refrigerator at 4°C, which are delivered on-demand via an automatic provisioning system. Cell culture material such as well plates and single-use pipet tips last for a minimum of four days until a restock is necessary. The refrigerator also holds media samples and stores them until further analysis. Dedicated software allows for a coherent and time-efficient input of all assay parameters as well as it provides a variety of tools for data analyzation. The Humimic LabOS also allows for assay feasibility checks and provides the operator with instructions for equipping the system. The fully automated image acquisition and ensuing analyzation will finally allow for higher comparability of results and new findings through pattern recognition and machine learning algorithms. Transferring our well-established co-culture assays with organ models such as liver, skin, intestine and bone marrow organoids proofed successful and showed high comparability to the manually conducted assays.
Novel Oxygen-Gradient Platform for the Co-Culture of Anaerobic Gut Microbiota with Primary Human Colon Epithelium
Raehyun Kim, University of North Carolina at Chapel Hill and North Carolina State University
Humans have co-evolved with their gut microbiota in a symbiotic relationship essential for health, yet how these thousands of bacterial species influence human biology remains little understood. A better understanding of the interplay between human cells and gut microbiota is required to exploit the complex relationships responsible for the local and distant effects of the microbiome on the human body. Accordingly, significant interest exists in the biotechnology community for improved in vitro models of the human gastrointestinal system, in particular models that support human-microbial co-culture. This feat is complicated by the fact that over 99% of gut bacteria are obligate anaerobes that die in 30-60 min after exposure to room air. Therefore, we have developed an easy-to-use and intuitive platform to replicate the steep O2 gradient across the in vivo colonic epithelium, thus create the appropriate environment required for anaerobes while maintaining viable, healthy epithelial tissue. We have computationally modeled, designed and prototyped the co-culture platform to fit within an SBS standard 12-well plate. The co-culture platform consisted of a basal reservoir and luminal reservoir with a porous polyester membrane and extracellular matrix (ECM) support dividing the two reservoirs. Using our culture methods, colonic epithelial stem cells were expanded on the ECM support and subsequently differentiated into all cell types found in the intestines in a monolayer ideal for compound screens and luminal stimulation/co-culture. Additionally, the ECM could be micro molded to recreate the physical architecture of the colon. Once the colonic epithelial layer was established, the luminal reservoir was sealed with an O2-impermeable barrier which resulted in the auto-generation of an anoxic environment (< 2% O2) in the luminal reservoir within 8 hours by the O2 consumption of the epithelial cells. The basal chamber remained normoxic to supply the epithelial cells with O2 through the porous membrane and ECM support. The generation of a steep O2 gradient was measured and experimentally confirmed. The resulting O2 gradient allowed for anaerobes (lactobacillus rhamnosus GG) to be cultured in the luminal reservoir in contact with an oxygenated colonic epithelial layer. Colonic epithelium and anaerobic bacteria each maintained >90% viability when co-cultured for ≥3 days. Our co-culture platform is simple, robust, self-sustaining and easy-to-use. It does not require any fluidic and gas control systems. It is based on a regular standard SBS microplate format that industry and academia use daily. Thus, it can be adopted in any microbiology laboratory without requiring new equipment.
Session Chair: Alejandro Amador, Ph.D., GlaxoSmithKline (USA)
Application of Genome-Wide Arrayed CRISPRn Screening for Target Discovery
Davide Gianni, AstraZeneca
Identification of novel and translatable therapeutic targets is urgently required for diseases with unmet clinical needs. Application of CRISPR/Cas9 technology in an arrayed screening format holds great potential for rapidly identifying new targets from physiologically relevant models. We have developed an end-to-end arrayed CRISPRn library screening platform for target discovery and integrated it with the capacity to interrogate advanced cell models. We are using this platform to perform gene perturbation experiments with the aim of identifying novel targets, understanding compound mode of action, building patient stratification hypotheses and ideas for novel combination therapies. A number of case studies will be presented highlighting the progress we have made in making this platform amenable to screening complex and advanced cell models. Taken together, these findings underline how functional genomics approaches using CRISPR/Cas9 technology are starting to revolutionize the drug discovery process.
Identification of novel and translatable therapeutic targets is urgently required for diseases with unmet clinical needs. Application of CRISPR/Cas9 technology in an arrayed screening format holds great potential for rapidly identifying new targets from physiologically relevant models. We have developed an end-to-end arrayed CRISPRn library screening platform for target discovery and integrated it with the capacity to interrogate advanced cell models. We are using this platform to perform gene perturbation experiments to identify novel targets, understanding compound mode of action, building patient stratification hypotheses and ideas for novel combination therapies. A number of case studies will be presented highlighting the progress we have made in making this platform amenable to screening complex and advanced cell models. Taken together, these findings underline how functional genomics approaches using CRISPR/Cas9 technology are starting to revolutionize the drug discovery process.
Presentation Title TBD
Using CRISPR-Cas9 Screening to Identify Genes Modulating the Nigericin-Induced Pyroptosis
Christian Parker, Novartis
Inflammasomes are multiprotein complexes that sense danger or damage-associated molecular patterns, DAMPS, as part of the innate immune system. This recognition leads to the release of cytokines and other signaling molecules that can then lead to cell death. Typically the term inflammasome refers to the complex of proteins including PYCARD (ASC), NLRP3 (NLRC4 or AIM2) and pro-caspase-1. Upon activation by various DAMPS this multi-protein complex promotes activation of caspase-1, which then leads to a cascade of events that cause the release of intercellular signals such as IL-β and IL-18. These danger signals, as well as released intracellular components, can then further activate inflammasomes present in surrounding cells. In addition, the activation of the inflammasome and caspase-1 can also lead to cell death due to the activation of membrane pores such as gasdermin D.
Mutations in components of the inflammasome have demonstrated this as a key pathway regulating autoimmune diseases; e.g. cryopyrin-associated periodic syndrome (CAPS), pyrin-associated autoinflammation with neutrophilic dermatosis (PAAND) and Familia Mediterranean Fever (FMF). A detailed understanding of the constituents of the inflammasome, and the pathway leading to its activation will have utility in designing treatments for a range of diseases associated with inflammation.
This report describes the development of an assay monitoring the induction of inflammasome mediated cell death (pyroptosis). The development of this assay allowed a genome-wide CRISPR-Cas9 screen to identify genes that regulate assembly and activation of the inflammasome. The assay utilized nigericin induction of the NLRP3 inflammasome mediated cell death in PMA differentiated THP-1 cells.
The screen successfully identified known components of the inflammasome as well as several genes that have not been previously implicated in inflammasome induced cell death. The use of a genome-wide screen allowed a comprehensive evaluation of the pathways controlling inflammasome assembly and activation. The top 1000 genes were identified for the creation of a focused mini-pool library of potential targets. Retesting of this mini-pool of potential targets confirmed the activity of many of these genes as modulating the inflammasome. So a further selection of genes was then made and these genes were knocked out individually using CRISPR-Cas9.
This presentation will discuss a number of the challenges faced invalidation of genes using this system as well as discussing potential means to address these issues.
Presentation Title TBD
Track Chairs: Yohann Potier, Ph.D., Voyager Therapeutics (USA) and Nicola Richmond, Ph.D. GlaxoSmithKline, (UK)
Session Chair: Umesh Katpally, Ph.D., Novartis Institutes for Biomedical Research (USA)
The Lab of the Future: Automation in the Digital Age
Michael Shanler, Gartner, Inc.
"Lab of the Future" (LoF) has recently become a popular topic for modernizing laboratories. While performing upgrades to laboratory informatics systems such as ELN or LIMS and adopting "hyped" technologies such as IoT, AI/ML, AR/VR and blockchain may support modernization on the surface, many existing LoF strategies run the risks of only delivering incremental value. Your LoF strategies need to have a deeper impact to survive executive scrutiny and must put an augmented data analytics strategy at the center. A LoF strategy must also enable a digital twin for the lab at multiple levels- with impacts on lab assets, personnel, and systems. As businesses transform, all aspects of business, including the laboratory need to support digital optimization and transformation. In this session, we review the meaning of digitalization, the technologies important for achieving LoF and outline strategic steps for ensuring your "LoF" strategy will be aligned to deliver true value in the Digital Era.
How Advances in Mobile, Voice and AI Technology are Impacting Scientists: The Evolution of Technology in the Lab - University of California, San Francisco Case Study
Ernesto Diaz Flores, University of California San Francisco
Scientists working today must navigate very large and complex datasets and work within regulatory boundaries that are tighter than ever. To meet the needs of modern scientists, lab documentation and management systems have had to evolve from simple pen and paper to flexible, integrated digital tools.
We are at an era in which technology is at our fingertips, and having new lab automation tools like a voice-powered AI digital lab assistant that allows integration of multiple functionalities within a laboratory annotation system greatly simplifies research workflows.
Ernesto Diaz-Flores is an Assistant Adjunct Professor at UCSF who works with his team of scientists in the lab to study novel therapeutic targets for high-risk subtypes of childhood leukemia. New technologies such as voice-powered AI digital assistants enable scientists at the UCSF lab to take voice notes, upload photos of experiments in real-time, set up several reminders throughout the day and even dictate what reagents they need in their shopping list, and have it all immediately added to their e-lab notebooks. There’s also been less human error, scientists can capture and access more information at the point of experimentation hands-free. When eyes and hands are occupied on the experiment, their voice can make the observations and capture the information in real-time with digital lab assistants.
As mobile, voice and AI technology evolve, there are now new options for scientists that seamlessly integrate with lab equipment and other data sources. The developments in mobile, voice and AI/machine learning technology are playing an important role in helping scientists bring their innovations and discoveries to market, improve efficiencies in the lab and make their work more reproducible.
How Advances in Mobile, Voice and AI Technology are Impacting Scientists: The Evolution of Technology in the Lab - University of California, San Francisco Case Study
Gursatya "Guru" Singh, LabTwin
Scientists working today must navigate very large and complex datasets and work within regulatory boundaries that are tighter than ever. In order to meet the needs of modern scientists, lab documentation and management systems have had to evolve from simple pen and paper to flexible, integrated digital tools.
We are at an era in which technology is at our fingertips, and having new lab automation tools like a voice-powered AI digital lab assistant that allows integration of multiple functionalities within a laboratory annotation system greatly simplifies research workflows.
Ernesto Diaz-Flores is an Assistant Adjunct Professor at UCSF who works with his team of scientists in the lab to study novel therapeutic targets for high-risk subtypes of childhood leukemia. New technologies such as voice-powered AI digital assistants enable scientists at the UCSF lab to take voice notes, upload photos of experiments in real-time, set up several reminders throughout the day and even dictate what reagents they need in their shopping list, and have it all immediately added to their e-lab notebooks. There’s also been less human error, scientists can capture and access more information at the point of experimentation hands-free. When eyes and hands are occupied on the experiment, their voice can make the observations and capture the information in real-time with digital lab assistants.
As mobile, voice and AI technology evolve, there are now new options for scientists that seamlessly integrate with lab equipment and other data sources. The developments in mobile, voice and AI/machine learning technology are playing an important role in helping scientists bring their innovations and discoveries to market, improve efficiencies in the lab and make their work more reproducible.
The Hyperloop for the Lab: An Integrated Approach for Sample Delivery and Treatment of Culture Dishes
Christoph Otto, TU Dresden
Digitalization, Automation and Miniaturization currently change the way we live and work. It also affects the daily work in laboratories creating what we perceive as the Lab 4.0 or the Lab of the Future. The disruptive development of new technologies such as open-source automation technology, the Internet of Things (IoT) and 3D-printing offer endless possibilities for the rapid engineering of new laboratory devices, which are compact, adaptable and smart. In conjunction with automated 3D-image analysis or deep learning algorithms, powerful instruments emerge to create and resolve research data. At the SmartLab systems department, the PetriJet31X hyperloop technology was developed to automate all processes associated with culture dishes in environments such as routine laboratories or culture development for the next generation of antibiotics. The device technically is an x-y-robot consisting of two linear axles enabled to transport all kinds of culture dishes from A to B through a 3D-printed gripper-system which can also remove the lid of the culture dish. The platform has been extended with a rail system and a small robot with the ability to transport piles of culture dishes or other laboratory material throughout the lab. The core part of the programming is a self-learning control software that does not need any teaching – the most time-consuming part of setting up a typical robot. With the presented solution an experiment conducted on samples is planned only once and executed for all culture dishes in the machine with the right processing station installed – e. g. 3D sample imaging and analysis. It is no longer necessary to specify locations for culture dish piles and treated dishes get allocated dynamically and drawn e.g. from the incubating chamber while user interactions are directed by LED-lighting. The system can process more than 1.200 culture dishes in an 8-hour shift and is equipped with a storage unit for these culture dishes. One example of the benefit of the PetriJet hyperloop can be found in routine labs for water and food inspection. Large numbers of samples get incubated on specific medium in culture dishes and are visually inspected regularly. Our system directly receives the tasks from the laboratory information and management system (LIMS), creates job lists and provides analytical data to the lab assistants through the LIMS. The data can then easily be turned into result sheets right from the desk. The unique feature of the system is that it can operate the night shift with no staff present. The PetriJet31X hyperloop platform now operates at the Chair of Microbiology at the TU Dresden for the screening of new antimicrobial substances and the next generation of antibiotics. The system enables biologists to screen agent combinations faster and use the gained image data to feed new deep-learning algorithms.
What, Where, How and Why? - Case Studies on Implementation of Lab of the Future Technologies in Discovery Life Sciences
Umesh Katpally, Novartis
There have been a good number of articles written, presentations made and webinars broadcasted on Lab of the Future (LoTF), including topics such as digitalization and automation, for the past few years. We continue to read, listen and discuss such topics even today and will continue to do so because the Future is always about looking ahead. In this presentation, we will look into the past few years to understand where we are currently in the context of implementing LoTF ideas and how they have been implemented. We will also look at what are some of the LoTF ideas that have not come to fruition and why. Based on this can we anticipate what new LoTF ideas will become into being in the years ahead?
Session Chair: Marti Head, Ph.D., Oak Ridge National Laboratory (USA)
Adventures in Data Sharing
Marti Head, ORNL
To democratize access to integrated data and transform our ability to make data-driven decisions, organizations must break down barriers between siloed data producers and foster a culture of open-hearted sharing of richly contextualized data. This talk will unearth adventures in cross-organizational data sharing, with storytelling and lessons learned over my years at GlaxoSmithKline Pharmaceuticals and Oak Ridge National Laboratory.
To Blockchain or not to Blockchain? Practical Applications, Benefits and Considerations of Blockchain Technology in Laboratory Workflows
Patrick Cullen, Yahara Software
In today’s world, no lab is an island. Samples, protocols, data, and ideas need to be shared seamlessly within organizations and across scientific communities to streamline the path to impactful progress. Also, the vast amounts of data generated for each sample can be overwhelming to analyze and maintain securely. Lastly, determining if data and results can be compared apples-to-apples between different protocols, studies, or organizations can be almost impossible.
In this talk, we will provide an overview of how blockchain can be a tool to achieve immense improvements in sample accessioning, data transfer, and workflow compliance using the ledgers inherent in the technology. Application of blockchain to these workflows can provide advantages such as elimination of ID duplication, reduction of missing or incorrect data, and decreased workflow errors and omissions, among others. When different organizations are part of the same blockchain, data integrity, protocol choice, and protocol adherence are transparent to all organizations involved. This shared technological ecosystem decreases barriers to communal data, study, and protocol repositories and increases the likelihood of productive collaboration.
Despite these benefits, the implementation of blockchain technologies is not without challenges, requiring new tools, strategic thinking, and updated approaches. To assist participants in evaluating opportunities for employing blockchain technologies within their own laboratories, we will provide an overview of some considerations organizations should take into account, offer practical steps to evaluate potential blockchain use cases, and discuss potential limitations of the technology.
Enabling "Bench-to-Bedside" with FAIR data
Viral Vyas, Bristol Myers Squibb
Translational Medicine (“bench-to-bedside”) is a multidisciplinary field focused on developing new therapies and procedures that extend and enhance human life. Collaboration with internal and external labs is at the very core of this effort which poses numerous challenges in fulfilling its promise. Biomarker data from thousands of patients and multiple indications need to be collected, collated with clinical data, analyzed and visualized. Data needed to gain insights is often hard to find, incomplete, opaque, lacks conformance and governance. Bristol-Myers Squibb has embarked on a set of bold strategic initiatives dubbed “Digital Health - Sage” to tackle these challenges head on by using FAIR (Findable, Accessible, Interoperable, Reusable) data principles as a guide. Five key capabilities were delivered as part of the Digital Health – Sage effort: data lake, data catalog, analytics environment, search and visualization, data access and governance. The presentation will detail business challenges, technology solutions, and lessons learned.
Agility, Security and Innovation - Lessons Learned While Developing One of the Largest Platforms for Analysis of Protected Healthcare Data
Edmon Begoli, Oak Ridge National Laboratory
In this talk, we will discuss our experiences gathered while developing one of the largest platforms for precision medicine, developed in support of the DOE-VA program MVP CHAMPION, and in this context, we will discuss the principles that guided the development of this unique platform. Specifically, we will focus on specific challenges and lessons learned while following the three core principles of the program - pervasive security, agility in all of the operations, and ongoing innovation. We will present these topics through the presentation of specific case cases and research scenarios.
Presentation Title TBD
Michael Boutros, German Cancer Research Center
Image-Based Cell Phenotyping Using Deep Learning
Samuel Berryman, University of British Columbia
The ability to phenotype cells is fundamentally important in biological research and medicine. Current methods of phenotyping cells rely primarily on flow cytometry to detect specific fluorescent markers. There are many situations where this approach is undesirable, such as problems with availability, specificity, cross-reactivity and cost of phenotyping markers. Furthermore, the number of markers required can increase the complexity, may exceed the detection limit, and even activate or decrease the viability of some cells. Finally, cells that are non-spherical or are too few are sometimes not compatible with flow cytometry. For these reasons, alternative methods for phenotyping are sought after, with the focus on live-cell imaging. Here, we investigate the potential to develop an “electronic eye” to phenotype cells directly from brightfield and non-specific fluorescence microscopy images.
Cells from ten cancer cell lines (MCF7, MDA-MB-231, LNCaP, PC3, U2OS, HCT, THP-1, HL60, Jurkat and Raji) were non-specifically stained to identify their nucleus (Hoechst), cytoplasm (Calcein green), and actin filaments (SiR actin). Cells were dispensed into 96-well glass-bottomed imaging plates and then imaged at 10X using brightfield and three fluorescence channels. The microscopy images were segmented into four-channel 51x51 pixel images each containing a single cell. The segmentation process used the DAPI channel for locating individual nuclei and then used a combination of the other channels to ensure other cells, or debris were not in close proximity.
To phenotype the cells, we developed a convolutional neural network (CNN) consisting of a four-channel input, four convolutional layers, one max-pooling layer, five fully connected layers, two dropout layers and seven batch normalization layers. ReLU was used as the activation function following each convolution or fully connected layer. Each node in the final fully-connected layer represented the probability of the imaged cell belonging to one of the known cell-lines. Softmax was utilized as a cross-entropy classifying error function for back-propagation during training.
Our CNN was trained over 20 epochs with Adam optimization and dropout to avoid overfitting and rotational data augmentation to expand the dataset. Using five-fold cross-validation, we show that the CNN was able to recognize each cell line with a 94% average accuracy. Our results demonstrate the ability to use deep-learning to phenotype cells directly from microscopy images without specific markers. This capability will be valuable for situations where phenotyping markers are unavailable or the cell sample cannot be stained (such as before therapeutic use). We envision this approach to be a general method for identifying cell types directly from image data to identify the emergence of phenotypic shifts or new cell types.
Assessing Biological Diversity of a Compound Collection Using High-Throughput Cellular Imaging and No Ground Truth
Yusuf Roohani, GlaxoSmithKline
Image-based profiling of cellular phenotypes has emerged as a powerful source of information for comparing chemical and genetic treatments. This has opened the door to interrogating the biological impact of a chemical collection at a scale that was not possible before. However, optimizing models to interpret these images generally requires ground truth labels for the mechanism of action that are difficult to generate at scale using conventional techniques. Most often what is available for compounds in the discovery stage is the nominal target but that does not capture primary and off-target effects. Moreover, plate, batch and instrument variation can complicate the transfer of methods and analyses across different datasets or instruments limiting the utility of public data for this purpose. Thus, research groups are compelled to build data analysis methods using the same instruments and protocols that they would apply to their own data, even in the absence of corresponding ground truth. In this talk, we describe a reliable system for discerning and labeling distinct image-based cellular phenotypes in such a scenario. We cover several methods for feature extraction (hand-engineered features, deep learning) and analysis (clustering, similarity metrics, hit calling, correcting batch effects). Most of our methods are based on the central principle that biological replicates exhibit identical phenotypes. We run a follow-up assays to validate our results.
Interpreting AI Models Trained on High-Content Microscopy Data
Oren Kraus, Phenomic AI
The increasing popularity of high-content screening (HCS) and phenotypic profiling in preclinical drug discovery is generating enormous amounts of complex imaging data. The flexibility of these imaging-based assays allows researchers to quantify many different biological processes using a single technology. Examples include examining nuclear translocation of proteins, internalization of receptors, and morphological changes in response to tens of thousands to hundreds of thousands of treatments. Despite the experimental throughput of HCS, analyzing and interpreting HCS imaging data remains a key bottleneck in utilizing these systems. Scientists often need to collaborate closely with computer vision experts and data scientists to extract informative measurements (i.e. features) from imaging data and design customized analysis pipelines for each new assay. Machine learning provides a unique opportunity to automate and accelerate many of the steps involved in analyzing HCS screens.
Recent results have shown that deep learning, specifically deep convolutional networks (CNNs) trained directly on raw pixel data, outperform existing approaches at classifying and clustering cellular phenotypes. The tradeoff often associated with these methods is the lack of interpretability of predictions made by deep learning models. We’ve designed several novel machine learning process for HCS that prioritize interpretability, by highlighting regions in the image that are responsible for the model’s predictions. These models combine fully convolutional neural networks, typically used for image segmentation, with convolutional multiple instance learning (convMIL) to aggregate predictions spatially across fields of view. Additionally, we’ve correlated predictions made by convMIL models with features extracted from individual cells using traditional feature extraction based analyses. Combining these two methods provides an additional layer of model interability by automatically indicating which features are changing most significantly between classes predicted by the CNN. Finally, we’ve developed a novel approach for exploring single-cell phenotypes in HCS screens using weakly-supervised learning models combined with an interactive tool for exploring phenotypes. Weakly supervised models are CNNs trained to predict every unique condition in an HCS screen based on image crops of single cells. Once the model is trained, a feature vector is extracted for every cell in the screen based on outputs from intermediate layers in the CNN. These feature vectors are then converted to 2D using dimensionality reduction techniques like t-SNE and UMAP. The interactive scatterplot we’ve built allows scientists to explore this 2D space while being able to see what individual cell phenotypes look like and which treatment conditions are common in different clusters that appear. We’ve used this tool to discover antibodies and compounds that are active in multiple assays that can include multiple cell-types and 3D culture systems. Taken together, these approaches significantly accelerate and improve phenotypic discovery programs.
Track Chairs: Margaret Scott, Ph.D., Genentech (USA) and Tim Wigle, Ph.D., Ribon Therapeutics (USA)
Session Chair: Robert Blake, Ph.D., Genentech (USA)
Leveraging Intrinsic Target Degradation
Delphine Collin, Cedilla Therapeutics
Protein degradation is a very effective way to inhibit target activity and prevent its scaffolding function. At Cedilla, we are focusing on small molecules that directly or indirectly regulate the homeostasis of a protein of interest. This presentation will cover some of our strategic approaches to direct and indirect degradation and the contribution of biomolecular sciences (biophysics, biochemical, structural biology, molecular dynamics…) to this novel path in drug discovery.
A Homogeneous Cell-Based Membrane Potential Assay to Identify Compounds That Promote Readthrough of Premature Termination Codons in the Cystic Fibrosis Transmembrane Conductance Regulator Ion Channel
Emery Smith, Scripps Research Institute Florida
Cystic fibrosis (CF), an inherited genetic disease, is caused by mutation of the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) gene, which encodes an ion channel involved in hydration maintenance via anion homeostasis. Nearly 5% of CF patients possess one or more copies of the G542X, which results in a stop codon at residue 542, preventing full-length CFTR protein synthesis. Identifying small molecule modulators of mutant CFTR biosynthesis that affect “readthrough” of this stop codon, or premature termination codon (PTC) to synthesize a fully functional CFTR protein represents a novel target area of drug discovery. We describe the implementation and integration for large scale screening of a homogeneous, miniaturized 1536-well functional G542X-CFTR readthrough assay. The assay utilizes HEK293 cells engineered to over-express the G542X-CFTR mutant, whose functional activity is monitored with a membrane potential dye. Cells are co-incubated with a CFTR amplifier and CFTR corrector to maximize mRNA levels and trafficking of CFTR, such that compounds that allow translational readthrough and synthesis of functional CFTR chloride channels will be reflected by changes in membrane potential in response to cAMP stimulation with forskolin, and CFTR channel potentiation with genistein. Assay statistics were excellent with Z’ values of 0.69±0.06 despite a S:B of 1.19±0.04. As further evidence of HTS suitability, we completed an automated screening of 666,120 compounds, identifying 7,761 initial hits. Following secondary and tertiary assays, we have identified 188 confirmed hit compounds with low and sub-micromolar potencies. Thus, the assay has integrated the advantages of a phenotypic screen with high throughput scalability to identify new small molecule G542X-CFTR readthrough modulators.
Monovalent Versus Bivalent Degraders
Robert Blake, Genentech
The emerging drug design strategy based on inducing target protein degradation offers the potential of drugging classes of proteins not previously thought to be druggable. Furthermore, the magnitude of effect for these agents is not limited by receptor occupancy and the duration of effects can persist beyond drug exposure. The current design of protein degraders is more commonly based on bivalent molecules, which consist of a ligand for the target protein linked to a ligand for a ubiquitin ligase (such as VHL, CRBN or XIAP). Due to their bivalent design, such drugs typically have a higher molecular weight than classic small molecule drugs and may present some non-ideal properties as drugs. An alternative monovalent degrader strategy is exemplified by the group of drugs termed SERDs (selective estrogen receptor degraders), for example, fulvestrant. The molecular structures of SERDs are typically designed around a receptor-ligand and a “degradation tail”, whose presence results in the degradation of the estrogen receptor. We have recently reported that this monovalent strategy can also be applied to the bromodomain and extra-terminal (BET) family, with the example of the monovalent BRD4 degrader GNE-0011. We will use examples of monovalent and bivalent degraders of BRD4 to compare these two complimentary degrader strategies.
Presentation Title TBD
Kandaswamy (Swamy) Vijayan, Plexium
Session Chair: Margaret Porter Scott, Ph.D., Genentech (USA)
Opening New Frontiers of Biology with RNA-Targeted Small Molecules
Jessica Friedman, Arrakis Therapeutics
The identification of drug-like small-molecule medicines that directly bind to RNA and modulate the biological function of RNA will vastly increase our therapeutic target space. RNA folds into structures that have diverse pockets into which small molecules can selectively bind. At Arrakis Therapeutics, our main focus is to identify small molecules that selectively bind RNA structures in critical regulatory regions of mRNA to modulate the expression of otherwise undruggable proteins. Our platform enables the analysis of RNA folding in vitro and in vivo as well as the identification of structures and pockets within an RNA. Drug-like compounds that bind these pockets are identified through high-throughput screening, followed by diverse biophysical assays to confirm and characterize binding such as SEC-MS, SPR and NMR. The biological impact of these compounds is evaluated in cell-based assays. The discovery of compounds that act via RNA binding is a new approach that has the potential to address previously intractable molecular targets and diseases.
Targeting Structurally and Functionally Diverse RNAs with Druglike Small Molecules
Jay Schneekloth, NIH
Recent estimates indicate that greater than 85% of the human genome is transcribed into RNA, yet just 3% of these transcripts code for protein sequences. Coupled with an increased knowledge of the noncoding functions of RNA and improved technologies for RNA structure determination, this information has given rise to interest in RNA as a therapeutic target for small molecules. In this presentation, I will discuss my group's recent efforts to understand RNA-small molecule interactions. We have developed a high throughput Small Molecule Microarray (SMM) screening platform, which we use to rapidly screen and profile RNA-binding small molecules. Further, I will discuss in detail several targets we have studied. I will also discuss our efforts toward the structure-guided design of RNA-binding molecules and the potential for future development.
Translating RNA Sequence into Lead Small Molecule Medicines and Progress Towards Small Molecule Antisense
Matthew Disney, The Scripps Research Institute
A major challenge in Medical Science has always been capturing targets for drug development. The state-of-the-art in targeting of RNA is the use of oligonucleotide-based modalities that target RNA sequence. Our focus over the past 15 years has been on developing technologies to decipher which cellular RNAs are “druggable” targets for small molecules and which small molecules can target them, serving as lead medicines. I will describe advances in the area of Small Molecules Interacting with RNA (SMIRNAs), including a sequence-based small molecule rational design tool dubbed Inforna. This approach allows sequence-based design principles for SMIRNAs of which only oligonucleotides have been previously designed from the sequence. Inforna has enabled the design of SMIRNAs against RNAs that cause hard to treat cancers and incurable genetically defined diseases that have no known treatment. We will describe these compounds and their implications for leveraging known biology to advance lead medicines and also their implications as chemical probes to understand previously unknown RNA biology.
I will also describe the development of approaches that allow for targeted degradation of RNAs in cells and animals by using SMIRNAs. For example, we have developed an approach that allows small molecules to recruit cellular nucleases to an RNA target to cleave it selectively and sub-stoichiometrically. Collectively, these studies show that small molecules can be designed to target RNA by using sequence-based design to deliver efficacious compounds targeting RNA including targeting the RNA for enzymatic destruction in cells and animals.
Development of a Novel High-Throughput Screening Approach to Target Specific RNAs
John Joslin, GNF
There is a growing appreciation for the many roles that structured RNAs play in disease progression. Accordingly, significant efforts are being devoted to identify low molecular weight compounds that specifically target these RNA molecules. These efforts may expand the number of therapeutic targets for disease intervention. While several groups are developing antisense oligonucleotides or siRNAs to target RNA, we sought to identify low molecular weight compounds that have desired drug-like properties, oral bioavailability, while retaining target specificity and potency. To accomplish this goal we have devised novel biochemical assays that are amenable to ultra-high-throughput screening. As a proof of concept, we have used these assays to identify compounds that bind well-characterized structured RNA, such as aptamers and ribozymes. For example, we developed screens that target the theophylline aptamer, which binds theophylline with high affinity and specificity. Through these screens, we identified theophylline, as well as several related molecules that bind the theophylline aptamer with an affinity that matches those reported in the literature. We also identified several compounds that are distinct from the theophylline scaffold, some of which have affinities exceeding that of the cognate ligand. To date, we have screened over 3 million compounds in these assays and have begun to understand the power and limitations of running HTS campaigns that target RNA. This presentation will highlight the assay development, screening results, and a view of what is required to carry out high-throughput screening campaigns that purposively target RNA.
Session Chair: Tim Wigle, Ph.D., Ribon Therapeutics (USA)
Re-Evaluating Kinase Inhibitor Selectivity and Residence Time in Living Cells with Energy Transfer
Matthew Robers, Promega
I will describe the application of an energy transfer technique (NanoBRET) that enables an approach to broadly profile compound fractional occupancy and residence time against a variety of target classes inside intact, living cells. Using this method, a broad-spectrum evaluation of compound engagement can be measured against over 300 full-length human kinases in live cells. Target engagement potencies correlate strongly with potencies using more traditional pathway analysis readouts, thus providing a platform to establish structure-activity relationships (SARs) for kinase chemical probes or lead drug molecules.
In live cells, we have uncovered a surprising spectrum of intracellular activity for certain cyclin-dependent kinase inhibitors (CDKi’s) and PROTACs, offering opportunities for repurposing some chemotypes as selective chemical probes for understudied kinases. We further evaluate opportunities for achieving target selectivity under non-equilibrium cell culture conditions, via protracted target residence time or PROTAC-mediated degradation. Here, we describe a broadly applicable approach for evaluating existing and novel chemical matter for selectively engaging CDKs in living cells.
A Bespoke Screening Platform to Study Mono (ADP-ribosylation)
Tim Wigle, Ribon Therapeutics
Mono(ADP-ribosylation) (MARylation) and poly(ADP-ribosylation) (PARylation) are post-translational modifications deposited on multiple amino acids, and emerging evidence suggests they are also deposited onto nucleic acids. There are 12 mono(ADP-ribose) polymerase (monoPARP) enzymes and 4 poly(ADP-ribose) polymerase (polyPARP) enzymes that use nicotinamide adenine dinucleotide (NAD+) as the ADP-ribose donating substrate to generate these modifications. While there are approved drugs and clinical trials on-going for inhibitors of the enzymes that deposit PARylation (specifically PARP1 and PARP2 inhibitors), MARylation is gaining recognition for its role in immune function, inflammation and cancer, however there is a lack of chemical probes to study the function of monoPARPs in cells and in vivo. An important first step to generating chemical probes for monoPARPs is to develop screening assays to enable determination of potency and selectivity of inhibitors during the hit finding and lead optimization phases. Complicating the development of enzyme assays is that the substrates for the majority of the monoPARPs are unknown, and even for the ones with identified substrates, it is uncertain how they engage their substrates. Here we describe the development of multiple family-wide approaches to developing robust high-throughput monoPARP assays that overcome this lack of knowledge around their substrates.
CETSA®-HT: Enabling a New Paradigm in Hit Discovery
Kirsten Tschapalda, AstraZeneca
High throughput screening (HTS) cascades have evolved to ensure that high-quality hits can be identified from large screening collections. Traditionally, most primary screens focus on the identification of modulators of catalytically active sites, while target engagement assays are placed further down the cascade. Well established technologies like competition-based assays, affinity selection technologies or differential scanning fluorimetry (DSF) depend on the availability of protein which is tested in a non-native biochemical setting. Therefore, one of the main concerns when initiating an HTS cascade remains the demonstration of target interaction within a relevant cellular environment. The use of cellular assays during primary screening and the HTS cascade presents an alternative. However, cell-based screens can easily become very complex, risk off-target effects and thus often require time-consuming target deconvolution of pathway hitters. To date, there has been no single technology that can demonstrate cellular target engagement in a suitable format for HTS primary screening. The cellular thermal shift assay (CETSA®) can act as an interface between this classic biochemical-cellular screening dichotomy. CETSA® facilitates label-free screening in disease-relevant cells while approaching the ease of biochemical assays. In an isothermal setup, full assay plates are heated to a set point within the target protein melting curve. While most proteins unfold and precipitate upon this heat-shock, a characteristic of protein-ligand interaction is induced thermal stability. The remaining stabilized protein can subsequently be detected with a pair of anti-species antibodies in an AlphaScreen® system. This high throughput (HT) CETSA® format allows large numbers of compounds to be tested in an HTS setting. Here, we report the development of two CETSA®-HT assays along with the application of this technology in HTS for the first time. This has been enabled following the recent agreement between Pelago Bioscience and PerkinElmer to streamline CETSA®-HT into validated kits and to offer support in the assay development. In the Global High Throughput Screening Centre of AstraZeneca, we are exploring the potential and the feasibility of CETSA®-HT for large scale HTS campaigns ( >0.5M compounds). These datasets indicate the future impact CETSA®-HT will have in hit identification. This is particularly timely given the expanding interest across drug discovery groups in new target protein classes. With new modalities like PROTACS (proteolysis targeting chimera) non-catalytically active proteins can now therapeutically be targeted. Utilizing CETSA®-HT to identify target engagement in cellular environments early during primary screening could shift the paradigm of hit finding.
Targeting Engagement by Utilising CETSA in Drug Target Studies
Laurence Arnold, Pelago Bioscience
Target engagement is a fundamental paradigm in drug discovery. A significant portion of projects fail to reach the clinic due to lack of efficacy or failure to show the lead candidate is interacting with the intended target in a more complex environment. Using the proven CEllular Thermal Shift Assay (CETSA) technology to measure target engagement in various matrixes, such as tissues, intact cells or lysates is increasingly common for SAR studies and lead optimization. It is also possible to inform early-stage programs with CETSA technologies and investigate in-situ ligandability. This study looks to investigate a well-established oncology target using CETSA high throughput platform (HT). CETSA HT measures direct target engagement through the interaction of the protein and molecule after a heat challenge, with versatile readouts applicable to HTS and miniaturization for robotic platforms. Assay development for CETSA HT includes the development of HTS off-the-shelf kit formats and in this project, investigations into target engagement of fragments within a cellular environment. These target engagement studies are not reliant on in-vitro and often abstract functional screens or methods, often limiting to certain protein classes.
Track Chairs: Amar Basu, Ph.D., Bioelectronica Corporation (USA) and Elodie Sollier, Ph.D., Benkei (France)
Session Chair: Alex Shalek, Ph.D., MIT (USA)
Enabling Precision Medicine: Pipetting at Single-Cell Resolution
Georges Muller, SEED Biosciences SA
Single-cell isolation is essential in stem cell biology, cancer research and biotechnology among others. For example, to ensure quality, safety and efficacy of the biotherapeutic product, companies shall demonstrate that each new recombinant cell line has been cloned from a single progenitor cell (WHO, 2014). Because available methods for cell cloning do not provide fully traceable cells yet, companies may waste up to 50 weeks in clonal validation. To solve this issue, we have developed an automated impedance-based pipetting robot for single-cell dispensing, allowing for traceable cloning of single cells. This technology permits the efficient and gentle isolation of industrial cell lines as well as rare and fragile stem cells and cancer cells, at a single-cell resolution, so that cells can be individually expanded in culture, transplanted downstream or analyzed by omics assays. We will present the technology and illustrate its key features through various case studies.
An Adaptable Microfluidic Platform for Single-Cell Pathogen Identification and Antimicrobial Susceptibility Testing
Pak Kin Wong, The Pennsylvania State University
Bacterial infections, such as bloodstream infections (BSI), ventilator-associated infection (VAI), and urinary tract infections (UTI), are a common cause of patient morbidity and mortality. Rapid identification of the causative pathogens and their antimicrobial susceptibility profiles will improve the clinical workflow for clinical management, accelerate clinical decision-making, and improve patient outcomes. However, definitive clinical microbiological analysis of samples obtained from patients requires several days, hindering proper management of infection and driving the overuse and misuse of broad-spectrum antibiotics. Novel precision technologies for rapidly identifying the pathogens and their antibiotic resistance are highly sought-after.
To address this clinical unmet need, we develop a nanotube assisted microwave electroporation (NAME) technique for intracellular detection of species-specific bacterial 16s rRNA in 30 minutes. NAME allows amplification-free pathogen identification at the single-cell level. Unlike typical sensing techniques that lyse the bacteria and dilute the intracellular content, NAME directly detects species-specific regions of the 16S rRNA inside the cells. Due to the small volume of a bacterium, the target molecule in the cell has a high effective concentration, which creates a strong signal for single-cell detection without amplification. Intracellular detection of bacterial 16S rRNA in viable cells also facilitates subsequent antimicrobial susceptibility testing (AST). By incorporating an adaptable microfluidic design, we demonstrate a phenotypic AST system that rapidly determines the existence of bacteria, classifies major classes of bacteria, detects polymicrobial samples, and identifies antimicrobial susceptibility directly from clinical samples at the single-cell level. The adaptable microfluidic system can dramatically accelerate the workflow of the microbiological analysis. Pathogen classification, which is based on microfluidic separation and microscopic inspection, eliminate the slow culture step. This approach rules out negative samples classify bacteria according to size and shape in as few as 5 minutes and identifies samples with multiple pathogens for polymicrobial infection diagnosis. By monitoring the bacterial growth directly, AST results can be reported in as few as 30 minutes or in a time scale similar to the doubling times of the bacteria.
In this study, we report the integrated microfluidic system for rapid pathogen classification and AST. We demonstrate the NAME technique for identifying bacteria that commonly cause BSI, VAI, and UTI. In collaboration with our clinical and industrial partners, we are developing an integrated ID-AST platform for the rapid diagnosis of bacterial infections. We pilot a study of 25 clinical urine samples to demonstrate the clinical applicability of the microfluidic system. The platform demonstrated a sensitivity of 100% and specificity of 83.33% for pathogen classification and achieved 100% concordance for AST. Our results demonstrate the analytical and clinical feasibilities of the integrated ID-AST platform for rapid microbiological analysis.
Building Tissue to Understand How Tissues Build Themselves
Zev Gartner, UCSF
The capacity of cells to self-organize into tissues is critical to their normal developmental and their ability to self-repair. Thus, a better understanding of how tissues self-organize will improve our ability to synthesize tissues and organs in the lab, and suggest new strategies to slow the breakdown of tissue structure that contributes to the initiation and progression of the disease. We are working to understand the mechanisms used by cells to self-organize robustly in the breast and gut, and how these programs are susceptible to the perturbations that underlie diseases such as cancer. We are also developing new single cell analysis technologies to help decode the logic of paracrine signaling networks that support the self-organization of these tissues.
Translating Single-Cell Genomics to the Clinic
Alex Shalek, Massachusetts Institute of Technology
While several methods exist for sampling tissues in clinical contexts, without high-fidelity tools for comprehensively profiling them, we are both limited in our capacity to understand how constituent cells and their interactions impact prognosis, and to select and develop precision therapeutics. Recent years have witnessed transformative and intersecting advances in nanofabrication and molecular biology that now enable deep profiling of low-input samples. Collectively, these afford new and exciting opportunities to study cellular heterogeneity, starting from the level of the single cell, and may unlock the diagnostic, prognostic, and discovery potential of clinical isolates. Illustratively, I will introduce how we can leverage single-cell genomic approaches – and, in particular, single-cell RNA-Seq – to explore the extensive functional diversity between cells, uncovering, from the “bottom-up,” distinct cell states and their molecular drivers. Moreover, I will discuss high-throughput experimental strategies and demonstrate how they can be leveraged to achieve the statistical power necessary to reconstruct intracellular circuits, enumerate and redefine cell states and types, and transform our understanding of cellular decision-making in health and disease on a genomic scale.
Session Chair: Daniel Austin, Ph.D., Brigham Young (USA)
Microscale Linear Ion Trap for Portable Mass Spectrometry
Daniel Austin, Brigham Young University
We present initial results from a high-aspect-ratio linear ion trap employing 20-micrometer-wide electrodes patterned onto ceramic substrates, with a characteristic trapping dimension of 800 micrometers. In previous efforts, we showed that a variety of ion trap geometries can be made using assemblies of two ceramic plates, the facing surfaces of which are patterned with appropriately shaped electrodes. The present report shows the significant miniaturization of this approach. Mass spectra of organic compounds with this device have a resolution of 2-3 amu. These highly miniaturized analyzers are now being developed for portable GC-MS instrumentation. Aluminum electrodes were deposited onto one side of each ceramic substrate. Electrodes are wire-bonded to a printed circuit board, which connects with a capacitive voltage divider. Two plate-PCB assemblies are mounted in a sandwich configuration, with the trapping fields being established in the space between the plates. Prior to patterning, a tapered ejection slit, 166 micrometers wide, was laser-cut into each substrate for ion ejection. The taper is critical to prevent ions from striking the inner wall of the slit and building up space-charge while allowing the thickness of the substrate to remain sufficiently thick for strength. Dipole resonant ejection of ions, in which the applied ejection waveform is phase-locked with the drive RF, was demonstrated by the use of special phase-tracking circuit. The alignment of the substrates was demonstrated using a set of 4 micropositioners (three linear and three angular). Low-power performance—essential for portable and hand-held mass spectrometers—was also demonstrated, with a maximum RF amplitude of 400 V at the highest point in the scan. The typical mass resolution of small organic compounds (toluene, xylenes) is 1.5 Da. Experiments using high molecular weight compounds (octofluorotoluene and perfluorotributylamine) showed typical mass resolution of 2-3 Da. The effects of higher operating pressure on mass spectra were also examined. Resolution decreased at pressures above 5 mTorr, but suitable spectra could still be obtained at pressures of up to 42 mTorr. Resolving power is decreased compared with the larger scale version of this device, possibly due to increased space charge. However, the signal to noise ratio is largely due to the high aspect ratio of these traps—the ratio of the length to the characteristic trapping dimension is greater than 40, providing a large trapping volume.
A Benchtop Biochemical Analyzer: Microchip Capillary Electrophoresis Coupled to High-Pressure Mass Spectrometry (HPMS)
J. Michael Ramsey, UNC Chapel Hill
Biochemical analysis needs are frequently addressed using liquid chromatography coupled to mass spectrometry (LCMS) in a centralized laboratory setting. While these systems can be quite versatile to address a broad range of biochemical measurement problems, they are correspondingly complex and require a trained operator to produce results. LCMS instrumentation also typically occupies a large footprint and requires utilities beyond a simple power outlet. Our laboratory has been pursuing miniaturized versions of liquid phase separation systems and mass spectrometers for over two decades. We are combining these two technologies to demonstrate a compact benchtop analyzer that can address measurement needs in areas such as cellular biology, clinical diagnostics, and biopharmaceutical research and development that would normally be accomplished using LCMS.
We have developed microfabricated capillary electrophoresis (microchip CE) devices with monolithically integrated nano-electrospray ionization (ESI) emitters that exceed the performance of conventional CE-ESI implementations. CE separations require ionic analytes, whereas LC can potentially separate either charged or neutral compounds. Biochemical species of interest are predominately ionic and CE systems outperform LC systems for separative performance, while the former can also be implemented more compactly with simpler components, e.g., voltage sources versus high-pressure pumps. Microchip CE has been used to separate ions as small as elemental species to intact monoclonal antibodies. One million theoretical plates of separation can be generated in one to a few minutes. Moreover, the microchip CE cartridge is easy to use and does not require any plumbing to connect the ESI emitter.
We have also been involved in the development of a new form of mass spectrometry, HPMS, that can be implemented in a compact form as it operates at pressures several orders of magnitude higher than conventional MS, i.e. approximately 1 Torr. Operating at such pressures allows significant simplification of the vacuum system and the use of a vacuum pump that can rest in the palm of your hand. The mass analyzer in HPMS is a form of ion trap with sub-mm scale critical dimensions. We have theoretically and experimentally demonstrated that HPMS resolution can be increased by decreasing critical dimensions and correspondingly increasing the RF drive frequency.
In this presentation, we will describe microchip CE and HPMS and the coupling of the two technologies to create a compact and useful biochemical analysis tool. The instrument implemented with a 96 well plate autosampler is approximately the size of a tower computer. Example applications such as monitoring bioreactor broth constituents will be presented.
A Small Footprint Ambient Ionization Enabled High-Throughput Chemical Detection System
Brian Musselman, IonSense, Inc.
Rapid analysis of the products of chemical reactions produced in high-throughput experiments (HTE) are completed by thermal desorption of sub-microliter volume samples into an ionizing gas. The heated ionizing gas completes the vaporization of the sample typically present in dimethyl-sulfoxide in 1-3 seconds per sample with rapid mass detection.The utility for Direct Analysis in Real Time (DART) for ionization of chemicals in the presence of aprotic solvents such as DMSO, and DMF has been employed to enable detection of those chemicals from sub-microliter volumes of the sample thus eliminating the need for sample dilution before analysis by LC/MS. The sub-microliter samples have been prepared by using several sample disposition method including low volume automated pipettor station and a high capacity disposable pin-tool. Using these devices we have chemicals present in concentrations appropriate for high throughput experiments that have been deposited onto a wire mesh surface which is then positioned between the DART source exit and the mass detector entrance for analysis. The use of small volume samples reduces the potential for matrix effiects by limiting the abundance of chemicals present in the ionizing gas. A rapid sampling of the small size droplets present on the sample supporting wire mesh enables continuous screening of the samples. We document the performance of each sampling method at analysis 1 per second to demonstrate the potential for a full 386-well sample plate analysis in under 10 minutes. Automated data analysis of the continuous collection of spectra in the data file is demonstrated using a file parsing software to permit the archival of the results. An outline of the overall workflow and its utility for simplifying the analytical effort in support of nanochemistry will be discussed.
ADE-OPP-MS: ESI-Mass Spectrometry-Based Bioanalytical Platform with Ultra-High Throughput
Hui Zhang, Pfizer Inc.
Recently, a new bioanalytical platform based on the coupling of acoustic droplet ejection (ADE) and open pore probe (OPP) technologies to mass spectrometry with standard electrospray ionization (ESI) ion source was introduced. Extraordinary performance has been demonstrated with this new platform, including plate-reader sampling speed, label free detection with MS, simplicity of assay development, just to name a few. As the pioneers to develop and apply this technology, our group has demonstrated the instrument capability and shared several seminal proof-of-concept studies supporting different areas of drug discovery including functional HTS, Drug-Drug Interaction (DDI), and other ADME applications in previous SLAS conference. With the commercialization of this new platform coming along, the technology had been further enhanced especially around system integration, automation, and robustness. We will be happy to report back the recent technology advancements, as well as the recent studies of the different screens Pfizer team has been working on to support live projects.
One lipid based HTS assay has been developed and successfully applied for hits triage and SAR support. Challenges of the lipid sampling and handling will be highlighted, and solutions and performance highlights will be provided. Both false positive and false negative hits identified from other HTS means were effectively teased out taking advantage of the highly selective of MS detection. With the integration of liquid handling workstation (Beckman I7) and other key peripherals to the ADE-OPP-MS instrument and implementation of automation solutions, ultra-high throughput screening with mass spec became a reality and can provide up to hundreds of thousands of compounds per day throughput. Besides pharmacology advancements, high throughput parallels medicinal chemistry applications enabled by this ADE-OPP-MS technology will be shared. Ultra-low sample requirements provide by ADE technology enabled readouts with as little as 1uL reaction, making 100X cost savings for the reagent and enabled much bigger chemical libraries to be made. Chemical reactions can be thoroughly evaluated with OPP-high resolution MS (HRMS) at second/reaction speed, 60-100X faster than the current paradigm. Such large and high-quality dataset enables the generation of big data around chemical reactions to feed in machine learning/artificial intelligence builds. Finally, biomarker analysis is another direction that we are interested in as another big advantage of the ADE-OPP-MS platform is that it can handle dirty sample matrixes with minimum sample preparation. We will show some recent example such as detecting N-methyl nicotinamide (NMN) from human urine samples or plasma pharmacokinetics (PK) studies where high-quality data (and equivalent to those obtained by conventional LC/MS method) were acquired without no sample preparation or just one step dilution; while the analysis time was effectively cut from multiple hours to a few minutes.
Session Chair: Élodie Sollier, Ph.D., Benkei (France)
Limitations and New Methods in the Characterization of Microfluidic Devices for Manufacturing QC
Maximilian Pitzek, Stratec Consumables
The range of applications for microfluidic devices is constantly expanding and so are the challenges for manufacturing. There is always exciting for the development of new manufacturing methods, but the importance of new analytical methods is often underestimated by contract manufacturers. Here we present our latest advances in the metrology and derived QC of injection molded microfluidic devices. Especially in the development of nm-sized fluidic channels, µm-sized fluidic channels with nm-precision, and devices with a combination of inorganic and organic coatings we came across challenges that required new analytical methods. The questions range from a simple "how deep is a microfluidic channel after bonding", to "what manufacturing process has the biggest impact on channel roughness - from Mastering to the finished device". To answer these questions, we had to find new ways to characterize the different processes along the manufacturing chain. The analytical methods that we developed opened up the door to new types of devices, or made it possible to mass manufacture new high complexity consumables.
A Portable Quality Control Lab in The Era of Food and Beverage Craftsmanship
Maciej Grajewski, SG Papertronics B.V. and University of Groningen
In the past decade, we have observed a tremendous growth of interest in food and beverage craftsmanship due to new trends for sustainability and eco-friendliness. This has resulted in numerous exciting initiatives in food and beverage production. However, these initiatives have limited resources and thus struggle to maintain consistently high-quality processing. Preventing losses, financial and otherwise, caused by unpredictable events such as microbiological contamination is often challenging.
Traditionally, these problems are tackled by quality-control (QC) protocols embedded in the production process. However, the implementation of QC often requires trained personnel and professional analytical equipment, with associated costs well beyond the budgets of small entrepreneurs. Therefore, it is a great opportunity for the rapidly growing field of portable (bio)chemical analysis to step in and offer a viable solution for this market. However, tests are often developed for trained chemists who are capable of proper sample acquisition and data interpretation. We propose an alternative technology that can accommodate common colorimetric tests, in a portable format that can be used by non-chemists. Moreover, the technology provides shorter sample-to-answer times for samples collected and analyzed directly at the production site by the craftsman. To achieve this, we utilize a patented sample concentration technology that enables quantitative analytical tests with enhanced sensitivity, combined with a sample acquisition system and data interpretation software.
The sample concentrator used in this work was developed by rapid prototyping with a stereolithographic 3D-printer. The concentrator cartridge consists of a sample acquisition module for volumetric sampling with subsequent liquid transfer to a porous particulate column packed into a 3D-printed cartridge. A series of branched air ducts embedded in the sample concentrator guide pressurized air from a simple gas supply to a selected region of porous membrane fixed into the bottom of the cartridge. Liquid samples are reacted with reagents in the 3D cartridge and then concentrated on the membrane through evaporation of solvent by the gas. The analysis is subsequently carried out by colorimetry. Integrated software analyzes the test results and compares it to results stored in our database so that every craftsman can use our system without the need for extensive chemical training. Additionally, our cartridges provide these tests with better protection against contamination, with improved user-friendliness, and the possibility of combining multiple materials for one test.
The described technology has been applied for QC testing in different branches of the food and beverage industry. Our technology contributes to the market of portable analysis because it provides a tool that can be applied outside an analytical laboratory, but with comparable results. This means shorter times between sampling and result, and, importantly, provides substantially better options for the small food-and-beverage entrepreneur to realize improved QC.
Single-Cell Hybrid Microfluidics as a Selection and Sorting Tool in the Mammalian Gene-Editing Pipeline
Kenza Samlali, Concordia University
Developing new engineered clonal cell lines is essential for loss-of-function studies, investigating protein function and unraveling signaling cascades and metabolic pathways. The current mammalian cell engineering pipeline of DNA delivery, selection, screening or sorting, and single-cell clonal expansion remains challenging and heavily relies on multiple expensive automated systems like a flow cytometer, cell sorters and colony pickers to increase experimental success.
Our group has shown the use of digital microfluidics for automating gene-editing procedures (Sinha et al.,2018), yet these devices cannot still select for successful edits or sort cells out into single clones.
Hybrid microfluidics combine the digital and droplet microfluidic paradigms in one device (Ahmadi et al., 2019), with electrode lined channels that can introduce more control over droplets in channels. In this talk, I will present four key results: First, we developed a hybrid microfluidic method for deterministic single-cell encapsulation. I’ll introduce a single-cell trapping array that is capable to achieve near-perfect one-cell-per-droplet encapsulation. Second, I will present the droplet operations that can be performed on this device. This includes precise on-demand droplet operations including releasing, merging and keeping single-cell containing droplets. This allows for dynamic assays of mammalian single cells, based on screening parameters that go beyond the traditional fluorescence based screening or sorting methods. All these operations, including encapsulation, can be easily performed with a graphical user interface. Third, I will describe the efficiency of cell trapping, cell encapsulation, droplet release and droplet keeping under different flow rates. Fourth, I will describe results that validate our platform for use in the mammalian cell engineering pipeline, using a breast cancer cell line (MCF-7) and a lung carcinoma cell line (NCI-H1299) as a model system. Specifically, I will show results related to encapsulation of a heterozygous and edited cell population and instant selection of transfected single-cells with subsequent generation of a clonal, isogenic edited cell line that could be expanded off chip.
This novel microfluidic device avoids the use of flow cytometry, cell sorters or limited dilution experiments to establish monoclonal engineered cell lines. Furthermore, the device can operate with low cell counts, which can potentially be used to handle sensitive cell lines along with enrichment of rare populations.
Microfluidic circuits with fluid walls to study cell migration
Cyril Deroy, Oxford University
A microfluidic technology based upon the use of “fluid walls” was recently introduced to address the lack of uptake of traditional microfluidic devices in biomedicine. Reasons cited for this lack of uptake include technical complexity, high failure rates due to gas-bubbles altering flows and affecting cells in micro-channels, the questionable bio-compatibility of materials like polydimethylsiloxane used to make devices, the inaccessibility of cells growing in them, and – probably the most important – biologists cannot use their familiar culture dishes and microscopes. The new technology reshapes fluid interfaces between two immiscible liquids (cell-growth medium and a bio-inert fluorocarbon, FC40) in a standard cell-culture dish, to form arrays of isolated liquid chambers. Each aqueous chamber is separated from its neighbors by transparent liquid walls of FC40. At the microscale, these fluid walls prove to be strong, pliant, and resilient. We now extend this fluid-shaping technology to create complicated circuits that can have almost any imaginable 2D shape, and demonstrate the power of the approach through a range of dynamic biological assays, in which cells migrate up concentration gradients established by diffusion. In one, primary mouse macrophages from the bone marrow are imaged as they migrate over polystyrene towards a chemo-attractant. More complex circuits in which cells are given a choice between different competing chemo-attractants allow the analysis of the decisions those cells make. In another, a steep gradient over less than 100 µm is generated by diffusion between two laminar streams as they flow through a conduit (width 500 µm, height 50 µm); bacterial cells (Pseudomonas Aeruginosa) growing in the conduit then migrate over glass towards an antibiotic. Finally, circuits with fluid-walls are built within wells of 96-well plates, to demonstrate that this technology can be seamlessly incorporated into standard high-throughput workflows based on microplates. All these circuits are built in minutes on virgin Petri dishes or standard plastics. The fluid walls allow direct access to any part of the circuit, they can even be reconfigured during the experiment, and cells can be recovered through them at any stage for analysis.
Track Chairs: Andrew Alt, Ph.D., University of Michigan (USA) and Guy Breitenbucher, Ph.D., University of California San Francisco (USA)
Session Chair: Guy Breitenbucher, Ph.D., University of California San Francisco (USA)
Integrated Phenotypic Screening and Chemical Proteomics to Discover Druggable Immunomodulatory Pathways
Ekaterina Vinogradova, The Scripps Research Institute
Modern human genetics has underscored the important role that the immune system plays in many human diseases ranging from autoimmunity to neurodegeneration to cancer. Nonetheless, chemical probes are still lacking for many immunologically relevant proteins and protein classes. We have applied an integrated phenotypic screening and chemical proteomic strategy to globally map the druggable proteome of human immune cells. Leveraging extensive past knowledge on the special capacity of cysteine-reactive small-molecule electrophiles to perturb immune system function, we have profiled this class of compounds against thousands of cysteines in primary human immune cells and will present compelling evidence that this strategy has already identified hit compounds for diverse classes of proteins with genetic links to human immune disorders.
Prestwick Chemical Innovative Screening Libraries To Support Drug Discovery
Christophe Morice, PC SAS - Prestwick Chemical
Screening is universally used to identify novel starting points in drug discovery and chemical biology programs. However, faced with a wide range of screening product offers, the selection of high-quality small molecule libraries remains a crucial step.
At Prestwick Chemical, we believe that coming and past success are strongly correlated and share Sir James Black’s statement when he says that “the most fruitful basis for the discovery of a new drug is to start with an old drug”. In this perspective, two valuable libraries based on marketed drugs have been developed at Prestwick.
In both libraries, molecules offer good coverage of the lead-like diversity space and present enhanced drug-likeness features by only bearing molecular functionalities present in already approved drugs.
Automated Chemistry: From an Idea to Reality
Michael Kossenjans, AstraZeneca
The discovery of bioactive small molecules is generally driven via iterative design-make-purify-test cycles. Automation is nowadays routinely used at the purify and test stage of these cycles but still very rarely at the design and make a stage. However, recent advances in areas such as microfluidics-assisted and batch chemical synthesis as well as AI systems that improve a design hypothesis through feedback analysis, are now providing a basis for the introduction of greater automation into all aspects of this process.
Here, we describe recent progress we have made with the build of a fully automated synthesis platform comprising all aspects of the make-purify workflow. A variety of chemistry transformations have been established on the platform allowing not only the rapid synthesis of compound libraries but also complex molecules via multistep sequences. The value and impact of our platform is illustrated in the context of specific case studies from the early phase drug discovery.
We also consider the longer-term goal of realizing the fully autonomous discovery of bioactive small molecules through the integration of our automated synthesis platform with automated design and testing.
Library Of Compounds with Really Annoying Pharmacology (LOCRAP)
Jarrod Walsh, AstraZeneca
False positive results have long been the bane of High Throughput Screening (HTS) campaigns. Depending on the compound library tested, the target itself and the assay system employed these have been estimated to account for enrichments of up to 95 % in hit outputs. Many sources of undesirable hits exist whether they be technology artifacts, redox cycling compounds, inhibitors of coupled enzyme systems or any other from a myriad of mechanisms. One strategy to cope has been the development of so called ‘nuisance compound’ sets. Over the past decade numerous pharmaceutical companies acting in isolation, including AstraZeneca, have generated their versions. Whilst the rationale, composition, source and application of these decks have varied between organizations the end goal has remained constant. Each group has aimed to either minimize the prevalence of undesirable actives by optimizing assay design or build bespoke triage cascades capable of identifying them. This presentation details the process by which the AstraZeneca set was established and shows the impact it’s had on drug discovery projects. We explore how it differs from those created by some of our peers and finally introduce a new cross industry and academia initiative called LOCRAP. This Library Of Compounds with Really Annoying Pharmacology (LOCRAP) is a collaboration between AstraZeneca, Eli Lilly, Novartis, Pfizer, Broad Institute, NCATS and lead academics in the area Jonathan Baell (Monash University) and Mike Walters (University of Minnesota). By pooling our collective knowledge, the intention is to design and deliver an industry standard collection that will be available to all through a third-party commercial partner. This new set will cover multiple classes of problematic compounds and contain thoroughly validated and well-annotated examples. We’ll discuss how this has been achieved and what classes are currently included. We are actively seeking contributors with ideas on compounds to include or to act as beta-testing supporters once the collection is available. We hope to make the tools and resources available to large pharmaceutical companies an option for any organization interested in drug discovery no matter how large or small. From this arises the aspiration to globally improve the quality of assays employed for hit identification and subsequently success rates for discovery campaigns.
Session Chair: Martin Matzuk, M.D., Ph.D., Baylor College of Medicine (USA)
Exploring Reliable and Cost-Effective DNA-Encoded Library Approaches for Developing Active Compounds
Raphael Franzini, University of Utah
Tagging combinatorial libraries with DNA barcodes allows using a simple affinity selection protocol to rapidly identify protein binders. A primary challenge with such DNA-encoded libraries (DELs) is how to design them to provide the effective screening productivity needed for the routine discovery of developable hits. The predominantly pursued approach is to make platforms of very large libraries of chemically complex compounds. While many success stories have proven the validity of this paradigm, it is unclear whether such DELs compare favorably to competing technologies with regard to return-on-investment. Moreover, the often heterogeneous synthesis yields of very large libraries and undersampling of DNA-barcodes impedes effective hit triaging. The cost of producing large DEL platforms and identifying hits from screening data is completely prohibitive for laboratories with limited resources. We, therefore, explore alternative library designs to find active compounds at lower DEL synthesis and lead-development costs. We custom-design DELs for specific target classes, and early studies have demonstrated that such libraries provide hits rapidly and economically. For example, a small and chemically simple DEL targeting NAD+-binding sites provided potent and target-selective hits for ADP-ribosyltransferases. Possible strategies for advancing early screening hits from such chemically simple libraries will be discussed.
Drug Discovery with DNA-Encoded Chemical Libraries
John Faver, Baylor College of Medicine
DNA-Encoded chemical Libraries (DEL) enable efficient screening of billions of drug-like small molecules for binding affinity to protein targets. With DEL, products of combinatorial chemical synthesis can be screened as a single mixture, because each compound is covalently linked to a DNA segment with a known identifying sequence. Affinity selection experiments are conducted to isolate small molecules that bind to protein targets under specific conditions, and DNA sequencing is used to identify binders and quantify enrichment. We have developed a DEL synthesis and screening platform in the Center for Drug Discovery at Baylor College of Medicine which utilizes novel on-DNA chemistries to generate both general and target-specific libraries. These libraries have produced hit series with robust structure-activity relationships for multiple target families. We have also developed enhanced data analysis strategies that allow for quantitative comparisons of enrichment from multiple screens, providing information such as target selectivity, locations of binding sites, and relative binding affinities. These strategies led us to the swift discovery of potent and selective hit compounds for a variety of targets with little additional medicinal chemistry optimization.
Off-DNA DNA Encoded Library Affinity Screening
Amber Hackler, The Scripps Research Institute
DNA-encoded library (DEL) technology is emerging as a key element of the small molecule discovery toolbox. DELs are highly diverse collections of combinatorially synthesized small molecules (~300-800 Da) attached to very large encoding DNA molecules. During affinity selection, the DNA tag can participate in the binding interaction, leading to false positive results and confounding the investigation of nucleic acid-binding targets (e.g. polymerases, transcription factors). An ideal affinity screen would only interrogate the library member for binding. Here, we use solid-phase DELs and microfluidic screening to separate each DEL member from its encoding tag and detect target binding using laser-induced fluorescence polarization (FP). DEL beads and an FP probe (dye-labeled ligand) are encapsulated in water-in-oil emulsion droplets containing the macromolecular target. Inside the droplet, the DEL member is photochemically cleaved from the DNA-encoded bead and laser-induced FP is measured. If the photocleaved library member competes with the probe for target binding, the probe emission is relatively depolarized, triggering electrokinetic droplet sorting and collection for follow up DNA sequencing. We prototyped this screening mode using the receptor tyrosine kinase (RTK) discoidin domain receptor 1 (DDR1), which is overexpressed in many cancers (e.g., leukemia, brain). A fluorescein-labeled DDR1 ligand (discovered in a previous affinity-based DEL screen) was used to build the droplet-scale competition binding assay. The FP signal difference between droplets containing either DDR1 or DDR1 and unlabeled competitor (20 µM) resulted in a Z’ of 0.58. After confirming assay robustness, a 67,100 member solid-phase DEL of drug-like small molecules was screened for ligands of DDR1 using the droplet-scale competition binding assay. Of the high-priority hit structures, several known RTK inhibitor pharmacophores were identified, including azaindole- and quinazoline-containing monomers. Off-DNA DEL affinity screening is amenable to screening in cell lysate and more complex affinity-based interrogations, such as interactome perturbation, in addition to providing an avenue to conduct mechanism-based screening using DEL.
DEL out of Water
Philip Dawson, Scripps Research
The structural diversity of DNA Encoded Libraries has been limited since the hydrophilic, unprotected nature of the DNA tag severely limits the repertoire of compatible chemical reactions. Rather than pursuing the optimization of individual synthetic organic reactions for water compatibility, we reasoned that a general strategy for transferring DNA-substrates into organic solvents could significantly expand the structural diversity explored by DEL. Reversible absorption of macromolecules to a solid support (RASS) has facilitated peptide and protein modification, enabling the use of anhydrous solvents and multistep synthetic procedures. This RASS strategy was adapted for DEL through a polystyrene based, quaternary ammonium resin. Adsorption of DNA headpiece substrates to this resin was found to facilitate transfer to organic solvents such as DMA, THF, and CH2Cl2. This RASS approach for DEL has enabled the development of Ni mediated carbon-carbon (C(sp2)-C(sp3)) and carbon-heteroatom (C-N, C-S, C-P) cross couplings with broad substrate scope and with excellent DNA compatibility. The immobilization of the DNA has also facilitated the use of electrochemical transformations. This expanded scope of reaction conditions compatible with DEL library generation has the promise to contribute to the generation of conformationally diverse scaffolds with drug-like properties.
Session Chair: Gerard Rosse, Ph.D., Dart Neuroscience (USA)
De Novo Drug Design with Chemistry-Savvy Machine Intelligence
Gisbert Schneider, ETH Zurich
Chemical creativity in the design of synthetic chemical entities with druglike properties has been the domain of medicinal chemists. At the same time, constructive machine learning models have been shown to autonomously sample drug-like molecules from chemical space without the need for explicit design rules. A machine learning method that combines a rule-based approach with a machine learning model was trained on synthetic routes described in chemical patent literature. This unique combination enables a balance between ligand-similarity based generation of innovative compounds by scaffold hopping and forward-synthetic feasibility of the designs. Prospective results demonstrate the capability of this hybrid machine learning model to capture implicit chemical knowledge from chemical reaction data and suggest feasible syntheses of new chemical matter. We will present various applications of molecular de novo design with machine intelligence, and discuss the advantages and limitations of these design concepts.
Button, A., Merk, D., Hiss, J. A., Schneider, G. Nat. Mach. Intell. 2019, 1, 307–315.
Schneider, G. Nat. Mach. Intell. 2019, 1, 128–130.
Schneider, G., Clark, D. Angew. Chem. Int. Ed. 2019, 58, 10792–10803.
Schneider, G. Nat. Rev. Drug Discov. 2018, 17, 97–113.
Discovering Multi-Target Pharmacology of Drugs and Drug Candidates by 3D Target Models and Machine Learning
Ruben Abagyan, University of California, San Diego
Small molecule therapeutics have an extensive, and only partially known, multi-target pharmacology that defines both their beneficial and adverse effects. We have characterized these networks for all cancer drugs. Furthermore, a team of researchers at Molsoft and UCSD developed a set of thousands of models in which 3D models are combined with the machine-learning layer to predict the activity of any chemical against protein targets included in this panel. The models can be used to discover targets of any known drug or drug candidate, search for compounds with specific multi-target profile, repurpose drugs for a new indication or disease, or identify potential liabilities. Applications of a multi-profile approach are presented.
Applying Artificial Intelligence and Machine Learning Techniques and Cross Platform Communication to Enable Informatics Driven Experimentation
Carleen Klumpp-Thomas, NIH/NCATS
Over the past decade, there has been a shift away from the traditional static method of performing high throughput screening (HTS) against large chemical libraries where experiments are designed in advance, executed and then the generated data is processed. Traditionally, this results in additional rounds of biological validation, testing and lead compound follow up for medicinal chemistry. More recently there has been a movement to focus instead on an increased number of targeted chemical libraries and smaller initial HTS experiments which can be run more dynamically with the resultant data processed automatically and in near real time to initiate new biological experimentation and even automated chemical synthesis on the fly. To make this possible it is necessary to have an underlying software and messaging infrastructure that can connect informatics platforms that utilize Artificial Intelligence and Machine Learning techniques to design experiments that can then be transferred to physical systems to initiate new experiments or small molecule synthesis. NCATS has developed such a platform with the initial validation being used to perform dynamic assay optimization but which is extensible to far more complex experimentation types to move beyond automation and instead towards autonomy.
Case Studies in AI-Driven Drug Design
John Griffin, Numerate, Inc.
Breakthroughs in machine learning theory and practice, coupled with ready access to cloud based supercomputing resources and ever-increasing amounts of experimental data, are enabling truly AI centric processes for small molecule drug design wherein predictive models successfully substitute for laboratory assays throughout the Discovery critical path. This presentation will describe how diverse applications of machine learning techniques, ranging from multidimensional and multitask boosting to deep neural networks, can extract accurate, scaffold independent, ligand based predictive models for important phenomena: target binding, functional activity, selectivity, PK/ADME properties, and toxicity. Applications of these models will be illustrated with examples from therapeutic programs and discussed in terms of their potential to enhance success/reduce attrition in drug discovery.
Track Chairs: Kristen Brennand, Ph.D., Mount Sinai School of Medicine (USA) and John Joslin, Ph.D., Genomics Institute of the Novartis Research Foundation (GNF) (USA)
Session Chair: John Joslin, The Genomics Institute of the Novartis Research Foundation (USA)
Toward Precision Medicines for Rare Diseases
Anna Greka, BWH/Harvard/Broad
Intracellular accumulation of misfolded proteins causes toxic proteinopathies, diseases without targeted therapies. Mucin 1 kidney disease (MKD) results from a frameshift mutation in the MUC1 gene (MUC1-fs). Here, we show that MKD is a toxic proteinopathy.
Intracellular MUC1-fs accumulation activated the ATF6 unfolded protein response (UPR) branch. We identified BRD4780, a small molecule that clears MUC1-fs from patient cells, from kidneys of knockin mice and from patient kidney organoids.
MUC1-fs is trapped in TMED9 cargo receptor-containing vesicles of the early secretory pathway. BRD4780 binds TMED9, releases MUC1-fs, and reroutes it for lysosomal degradation, an effect phenocopied by TMED9 deletion. Our findings reveal BRD4780 as a promising lead for the treatment of MKD and other toxic proteinopathies. Generally, we elucidate a novel mechanism for the entrapment of misfolded proteins by cargo receptors and a strategy for their release and anterograde trafficking to the lysosome.
A Hydrogel-Enabled 3D Liver Fibrosis Model for High-Throughput Phenotypic Screening Applications
Zhixiang Tong, GNF/NIBR
Hepatic stellate cell (HSC) is one of the major cell types responsible for the progression of liver fibrosis. Inhibition of HSC activation (i.e. transdifferentiation from an inactive/quiescent state to a myofibroblast-like state) represents a compelling strategy for resolving liver fibrosis. Conventional 2D HSC culture cannot recapitulate the in-vivo 3D cell-ECM interactions and often causes undesirable cell activation due to the intrinsic mechanical properties of the plastic substrates. Hydrogel has emerged as a powerful tool for modeling the 3D cellular microenvironment in vitro, largely due to its chemical and physical versatility. To this end, GNF team has developed a comprehensive toolbox comprised of a variety of polyethylene, hyaluronic acid and gelatin-based hydrogel derivatives. Specifically for modeling HSC activation, we’ve identified a gelatin methacrylate (GelMA)-based hydrogel formulation that preserves the quiescent state of HSCs much better than 2D culture on tissue culture polystyrene, and precisely controls the pro-fibrotic signal (i.e. TGFβ) induced activation in 3D. Such control offers optimal TGFβ induced response window superior to that of standard 2D culture, leading to robust assay performance in a highly miniaturized format. By leveraging GNF’s proprietary engineering/automation technologies, the 3D culture can be efficiently implemented in a 1536-well plate format with minimal inter-/intra-plate variations. Using a tool GNF compound set (with ~1800 compounds), our pilot high-content imaging (HCI) based phenotypic screen reveals an assay robust Z’ factor over 0.45 and a hit-picking rate ~3%. Overall, our automation-friendly, hydrogel-enabled 3D HSC assay can be readily scaled up either for primary screen or lead profiling applications, and the novel screening platform shown herein could offer broad utility to identifying novel targets or therapeutic modalities for diseases beyond liver fibrosis.
Development of Colorectal Cancer Patient-Derived Organoid-Fibroblast Models Suitable for Drug Testing
Eliza Fong, National University of Singapore
The advent of patient-derived organoid (PDO) technologies has greatly expanded the toolbox for drug discovery and personalized drug screening for several cancer types. While PDO models more closely represent the molecular characteristics and heterogeneity of patient tumors than traditional immortalized cancer cell lines, they are inherently limited in their ability to reflect the tumor microenvironment in vitro as they comprise exclusively of epithelial cells. The lack of stromal cells in PDO models, such as cancer-associated fibroblasts (CAFs), poses a major problem as these tumor microenvironmental components contribute to the various hallmarks of cancer and response to therapy. Particularly for colorectal cancer, CAFs comprise the majority of the tumor microenvironment and play important roles in cancer development and progression, from the regulation of cancer cell proliferation and stem cell maintenance to drug resistance. In this study, we addressed this problem by establishing in vitro conditions that robustly enable the co-culture of CRC PDO with patient-derived CAFs for controlled mechanistic studies and drug testing. We report the development of an engineered tumor microenvironment consisting of CRC PDO encapsulated within a well-defined three-dimensional (3D) hyaluronan-gelatin hydrogel and co-cultured with patient-derived CAFs. Basement membrane extracts (e.g. Matrigel) conventionally used for PDO culture exhibit batch-to-batch variability. Considering that the CRC extracellular matrix is high in hyaluronan and collagen I and that hyaluronan-based matrices are conducive for the culture of various human cancers, we hypothesized that hyaluronan-gelatin hydrogels may serve as a suitable alternative 3D matrix to support the culture of CRC PDO and CAFs. Through RNA- and whole-exome sequencing, we first show that these hydrogels are capable of maintaining the molecular characteristics of the original patient tumors in the cultured CRC PDO. Further, based on our findings that standard PDO culture medium poorly supports CAF viability, we developed a new co-culture strategy that robustly maintains the viability of both CRC PDO and CAFs for at least a week in culture. We found that in the absence of any growth supplements added to the co-culture, CAFs were able to maintain the growth of the cultured CRC PDO in the hydrogels. Lastly, we demonstrate that these CRC PDO-CAFs models are suitable for evaluating standard-of-care drugs, making them potentially very useful for realizing personalized cancer medicine.
Liquid Biopsy as a Tool in Ant-Cancer Drug Screening for Personalized Therapy Decision and Drug Discovery
Kamran Honarnejad, Fraunhofer ITEM
The use of circulating tumor cells (CTCs) isolated from liquid biopsy are already used to predict disease progression and survival in metastatic patients. However, the lack of robust drug screening assays has hampered their application in monitoring patient drug response/resistance and personalized therapy decision.
We have developed a workflow to isolate tumor cells from pleural effusion and malignant ascites samples from metastatic lung and breast cancer patients and subjected them to medium scale drug screens against approved anticancer drug libraries. This approach allows the realization of personalized treatment decisions within less than a week by evaluating drug responses directly in patient-derived tumor cells obtained from liquid biopsy.
In patients with no pleural effusion or malignant ascites, we have established another workflow to isolate viable CTCs from peripheral blood of metastatic patients, from which we have generated 2D and 3D in vitro (spheroids and organoids) and in vivo (CTC-derived xenografts) models. Particularly, drug screens on CTC-derived organoids were feasible within therapeutic timeframes which can potentially influence personalized treatment strategy. Drug responses from the screen mirrored patients’ drug resistance and revealed promising candidates for the treatment of individual patients. Beyond that, high-throughput drug screens in CTC-derived preclinical models closely mimicking patients' settings enable discovery, repurposing and development of more efficient cancer therapeutics.
Integration of drug screening of liquid biopsy-derived tumor cells constitutes a powerful tool to better improve personalized treatment strategies and discovery for metastastic patients.
Session Chair: Kevin Eggan, Ph.D., Harvard University (USA)
Presentation Title TBD
Kevin Eggan, Harvard University
Developing Epileptic Encephalopathy Models Using iPSC-Based Technologies
Evangelos Kiskinis, Northwestern University - Feinberg School of Medicine
Mutations in KCNQ2, which encodes a pore-forming K+channel subunit responsible for neuronal M-current, cause neonatal epileptic encephalopathy, a complex disorder presenting with severe early-onset seizures and impaired neurodevelopment. The condition is exceptionally difficult to treat, partially because the effects of KCNQ2mutations on the development and function of human neurons are unknown. Here, we used induced pluripotent stem cells and gene editing to establish a disease model and measured the functional properties of patient-derived neurons using electrophysiological and optical approaches at single-cell resolution. We find that while patient-derived excitatory neurons exhibit reduced M-current early, they develop intrinsic and network hyperexcitability progressively. This hyperexcitability is associated with faster action potential repolarization, larger afterhyperpolarization, and a functional enhancement of Ca2+-activated K+(BK and SK) channels. These properties facilitate a burst-suppression firing pattern that is reminiscent of the interictal electroencephalography pattern in patients. Importantly, we were able to phenocopy these excitability features in control neurons only by chronic but not acute pharmacological inhibition of M-current. Our findings suggest that dyshomeostatic mechanisms compound KCNQ2 loss-of-function and lead to alterations in the neurodevelopmental trajectory of patient-derived neurons. Our work has therapeutic implications in explaining why KCNQ2 agonists are not beneficial unless started at an early disease stage.
Application of Microwell Plates in Single-Cell Analysis of T Cell-Mediated Tumor Cell Killing for High-Throughput Pharmacological Analyses
Katherine Kozak, Genentech
The ability to observe and quantitate T cell-mediated tumor cell killing at the individual cell level is critical for understanding the mechanism of immune activation and exhaustion to assist therapeutic designs. To acquire a large amount of single-cell data for statistical analysis, images of cells in micro-gridded chambers (sub-wells) within a standard microwell are captured to analyze single-cell interactions. The current available gridded platforms, manual microscopy imaging or standard automated image cytometer methods, however, are time-consuming. In addition, the analysis software is typically used for custom-made sub-wells constructed with Polydimethylsiloxane. In this work, we reported a high-throughput single T cell killing assay utilizing the Celigo Image Cytometer and Elplasia SQ plates that convert each 384-well into 86,400 sub-wells. In this assay, both T cells and tumor cells were seeded with varying densities, resulting in 1 to 100 cells in each sub-well, as well as varying T cell activation reagent. The plate was scanned immediately after cell seeding at t = 0, 4, 22, 46, 68, and 168 hours, where the number of cells were tracked in each individual sub-well over time for each cell type. The image cytometer was able to rapidly acquire and analyze images at 1 µm2/pixel. The images and results were exported into a custom program to determine the proper sub-well location of the segmented cells to allow separate tracking of small groups of tumor and T cells. This approach gave the image cytometry method greater capability in resolving subpopulations within the biological sample, resulting in more detail on the cytotoxic killing as a function of cell demographics in the tumor microenvironment.
Microfluidic Platform for Screening of Antibiotic Susceptibility at the Single-Cell Level
Witold Postek, Institute of Physical Chemistry of the Polish Academy of Sciences
The inoculum effect describes a dependency between the minimum inhibitory concentration (MIC) of an antibiotic and the concentration of bacteria in the sample: the less the bacteria, the less concentrated antibiotic is needed to stop their growth. MIC for populations consisting of a single cell is known as single-cell MIC (scMIC). scMIC is important for public health, as the presence of antibiotics at a concentration of scMIC in a large population of bacteria drives the evolutionary pressure towards resistant strains1, and the inoculum effect is a source of errors in MIC assessment in the clinic. However, efficient assessment of scMIC values for large numbers of cells has not been shown until now.
Here, we demonstrate a method of determining scMIC values in hundreds of replications per experimental run, and we achieve this without optical labeling of the reaction conditions. We generate a series of emulsions of different concentrations of antibiotics at a step emulsifier2. We encapsulate single cells in each emulsion droplet due to stochastic confinement. Each emulsion is separated from the others by being encapsulated in a third immiscible phase and transferred to a piece of tubing, where all the separated emulsions can be incubated to provide for growth of bacteria. We measured the scMIC value of cefotaxime in E. coli for hundreds of cells, recording the inoculum effect when we used higher initial cell densities and observing the distribution of resistance level in a population of bacteria. Currently, we use our platform to generate up to 20 separate emulsions with different and known reaction conditions of ca. 2000 droplets each with immediate plans to upscale. In the near future we plan to screen for interactions of antibiotics in relation to inoculum effect, including the measurements at the single-cell level.
The described method might be useful in the field of antibiotic resistance at a single-cell level, which is unbiased by the inoculum density. A microfluidic method of screening multiple chemical conditions in emulsions without labeling can be also deployed in other fields of research, wherever several reaction conditions should be replicated hundreds or thousands of times. For now3, to establish whether the bacteria grow or not, we detect fluorescence from fluorescent proteins produced by bacteria, but we are currently working on an add-on module to detect growth without labelling. We are also integrating our system with optical detection of moving droplets to automate the liquid handling protocol.
Session Chair: Kelly Frazer, Ph.D., University of California San Diego (USA)
Overlap of Fetal-Specific Cardiac Regulatory Variants and GWAS Lead Variants Supports Fetal Origins of Cardiovascular Disease
Kelly Frazer, University of California, San Diego
It has been hypothesized that many disease-causing variants exert their effects during development, rather than in adult cells. However, it is difficult to identify these variants and their effects as they could act in multiple different cell types, and there was a recent moratorium on research using fetal tissue. We recently established that iPSC-derived cardiovascular progenitor cells (CVPCs) are fetal-like, and can be utilized to identify cardiac regulatory variants. Here, we leveraged this system to identify fetal cell-type-specific eQTLs that underlie GWAS signals for adult cardiac diseases. We started by characterizing the differentiation of iPSCs into iPSC-CVPCs via scRNA-seq on eight samples, and found they were comprised of two cardiac cell types: cardiomyocytes (CMs) and epicardium derived cells (EPDCs). Next, we derived 180 iPSC-CVPCs, performed bulk RNA-seq, and used the scRNA-seq expression signatures to deconvolute and determine the relative proportions of CMs and EPDCs in each sample. We integrated these data with WGS and identified cell type-specific eQTLs (associated with only CMs or EPDCs). We next identified fetal-specific eQTLs by colocalizing our iPSC-CVPC eQTLs with all GTEx adult cardiac tissue eQTLs. To identify variants underlying the fetal origin of complex adult cardiac traits, we colocalized these fetal-specific eQTLs with cardiac traits GWAS summary statistics (pulse rate and myocardial infarction) and found 10 fetal-specific eGenes, including CLPTM1 which has previously been associated with congenital malformations (as expected for a fetal-acting gene). Our findings provide genetic evidence supporting the fetal origin of cardiovascular disease and show that iPSC-derived tissues can be leveraged to study the fetal origins of diseases in relevant cell-types.
Presentation Title TBD
Paul de Bakker, Vertex
CURATE.AI-Enabled Personalised Dosing for Multiple Myeloma
Agata Blasiak, National University of Singapore
Standard of care therapy for various indications, including multiple myeloma (MM), is a combination of up to 4 drugs. Precision medicine has emerged as a game-changing approach for selecting drug combinations tailored to an individual to avoid treatment failure in a common case of drug resistance. To maximize the therapeutic outcome with the identified combination therapy, the drug dosing strategy should undergo an analogical approach - personalized dose selection for each drug tailored to an individual. Conventional approaches - titration, additive drug design, and dose escalation – often fail to find the optimal doses. Modern approaches - predictive algorithms and genotypic modeling – require a substantial amount of data and are costly. In this pilot study, we use CURATE.AI, a disease mechanism-independent and indication agnostic platform, to create an N-of-1 drug interaction profile using only the patient’s own data to identify optimized doses. CURATE.AI has been already clinically validated and has been used to optimize combination therapy for acute lymphoblastic leukemia, combination therapy for prostate cancer, liver transplant immunosuppression, and tuberculosis therapy, among other indications. We applied CURATE.AI to retrospectively analyze medical dataset in accordance with institutional IRB. As indicated in the medical records, a patient was given 14 monthly modulated dosages of revlimid and cyclophosphamide and a constant monthly dosage of dexamethasone. Quadratic polynomial correlation between the drugs’ dosages and the platelet count - a clinical indicator of the disease progression - was used to create the patient-specific CURATE.AI profile, which served as a map to identify drug dosages within clinically-accepted ranges that would result in an optimum platelet count. Using CURATE.AI analysis, the platelet count (P(r,c)) was correlated to revlimid and cyclophosphamide concentrations (r and c, respectively) by the following function: P(r,c)=62+4.690r-0.069rc+0.275r2+0.002c2, with R2value of 0.724 and a fitting correlation of 0.851. The CURATE.AI profile guided that to sustain the platelet count within the desired range (132-372x109/L), the patient should be given above 10 mg matched with c below 150 mg, or cabove180 mg matched with below 15 mg. CURATE.AI is deterministic and does not involve any prediction or uncertainty of response. In addition, CURATE.AI recommends doses within clinically-accepted ranges at patient-specific time points to optimize treatment response for that particular patient. The ability to identify patient-specific response constants has a paradigm-shifting potential – combining precision medicine for drug selection with CURATE.AI for dose selection brings us a step closer to the true realization of personalized medicine. We have also initiated a clinical trial that uses CURATE.AI for prospective dosing in MM (Clinicaltrials.gov: NCT03759093).
Personalizing Digital Therapeutics with CURATE.AI Identified N.1 Profiles
Theodore Kee, National University of Singapore
Digital therapeutics have emerged as an alternative or complementary modality of treatment to drug-based therapies for various indications, such as addiction and cognitive decline. Similar to conventional drug dosing, digital therapies often rely upon either fixed, or step-wise increased difficulty. Those approaches lack the flexibility and the ability to personalize the treatment to the individual. URATE.AI - a clinically validated deterministic optimization artificial intelligence (AI) platform that has already been used to modulate optimized dosing regimens for indications ranging from oncology (solid tumor/hematologic) to infectious diseases (HIV/TB) and immunosuppression (liver). In this prospective study, CURATE.AI identified individualized N-of-1 (N.1) learning trajectory profiles of healthy volunteers (both sexes, ages 21-40) trained on the Multi-Attribute Task Battery (MATB). MATB is a flight deck simulator developed by the National Aeronautics and Space Administration (NASA) and United States Air Force (USAF). The prospective clinical trial study design is a randomized, multiphase, parallel three-arm, single-blinded, N-of-1, single-center, exploratory pilot trial with 1:1:1 allocation approved by NUS Institutional Review Board (S-17-180) and listed under Clinicaltrials.gov identifier NCT03832101. For the CURATE.AI arm of the study, five subjects were fluent in English, had no prior experience with the MATB, and no history of perceptual or memory deficits, and recruited at Yale-NUS to participate in MATB simulator experiment sessions, conducted at the Yale-NUS campus. Each subject underwent a 34-minute MATB training session composed of 17 training and testing blocks of varying intensity levels (high, medium, low). Individualized CURATE.AI profiles were calibrated from the individual’s data: performance scores (RMAN-COMM z-scores from the training blocks), performance improvement, and training intensity levels. From each individual’s prospectively obtained data, N.1 learning trajectory profiles were derived and constructed with CURATE.AI, demonstrating the unique relationship between performance, training intensity, and performance improvement. Each subject had a different performance range (-1.24 to 0.70, -1.59 to 1.04, -2.13 to 0.66) and different training intensity levels for optimal performance improvement. As identified from their CURATE.AI N.1 profiles, high-intensity training in select participants corresponded with greatest gains in performance improvement, while low-intensity training was identified for mediating similar gains in the other subjects. Based upon each individual’s unique interaction between performance and training intensity, these N.1 learning trajectory profiles provide a means of real-time optimization of performance improvement by dynamically identifying and modulating the appropriate training intensities. In this prospective in-human study, interfacing MATB with CURATE.AI revealed substantial differences between subjects’ N.1 learning profiles and the correlation between tailored training intensity on performance improvement. The ability of CURATE.AI to identify N.1 profiles represents the advancement and utilization of AI to actionably address challenges encountered in personalized learning and the emerging field of digital therapeutics.