SBI 2 High Content 2016,3 rd Annual Conference,September 12 th -14 th

Biology that unfolds slowly, asynchronously or in a cell-specific manner affecting only a subpopulation of cells can be challenging to study. Conventional approaches rely on cross-sectional comparisons between unrelated cells made at single time points. Cell-to-cell differences often are treated as technical rather than biological variation, significantly reducing the sensitivity of the approach to uncover meaningful biological differences or the effects of genetic or small molecule perturbagens.

To overcome these and other confounds, we developed systems to perform high throughput longitudinal single cell imaging and analysis that we call robotic microscopes. A robotic incubator delivers a user-specified microtiter plate to an external nest, a robotic arm transfers the plate to the stage of a fully automated microscope stage, and the microscope is programmed to perform automated image acquisition. The entire system sits within an environmental chamber and operates around the clock without human intervention. Critically, software programs enable the microscope to return to exactly the same cell as often and for as long as the user wishes, while automatically returning the plate to the incubator between imaging episodes. Automated programs find cells within the images, assign them unique identifier numbers and track them longitudinally. Increasingly, deep learning approaches are used to extract features from images. The responses of thousands of individual cells are compared with time-to-event or survival analysis. Longitudinal single cell analysis is 2-to-3 orders of magnitude more sensitive than cross-sectional comparisons of unrelated cells. The exquisite sensitivity arises in part because each cell effectively serves as its own control, which significantly reduces variability, and because of the cumulative nature of the biological measurements is well suited to slowly unfolding, asynchronous events. These strengths make the approach particularly well-suited for small molecule or genetic screening of heterogeneous or precious cell populations, such as primary culture, differentiated human induced pluripotent cell culture, organotypic tissue or even model organisms.

10:10–12:30 PM  SBI² Session I: Image Based HCS

Success in high throughput screening depends on developing a robust and consistent biochemical or cellular assay. Image-based assays have reached these standards and therefore enable highly sophisticated assay endpoints including the translocation of transcription factors, cellular morphology and colony behavior. This session will highlight examples of such cellular events as screening endpoints for chemical biology and therapeutic development.

Session Chairs:

Vance Lemmon (Centre for Computational Science, University of Miami)

Ann Hoffman (NCE Drug Design and Selection, Glaxo Smith Kline)

10:10–10:40 AM  Invited Speaker

Ras Imaging and High Content Assay Development

Turbyville Thomas NCI Frederick

Mutations in RAS drive 30% of human cancers. In order to carry out its pathogenic functions, mutated RAS must localize to the plasma membrane of cancer cells where it associates with downstream effector molecules. One strategy to target mutant RAS, therefore, is to disrupt its association with the plasma membrane. We have developed a high content assay to identify such molecules, and have developed a number of complementary approaches to characterize RAS interactions and behavior in the plasma membrane.

10:40–11:10 AM  Break

11:10–11:40 AM  Invited Speaker

Life beyond the pixels: Genome-wide phenotypic discovery using machine learning and image analysis methods

Horvath Peter Biological Research Centre of the Hungarian Academy of Sciences

This talk gives a summary of good and bad experiences collected after the multi-parametric analysis of several human genome-scale RNAi experiments. A comparative study between different siRNA libraries will be presented. Finally novel image analysis and weakly supervised machine learning techniques will be discussed to rapidly and precisely analyze high content data.

11:40–12:30 PM  Two short talks selected from submitted abstracts

12:45–1:30 PM  Technology Spotlight Lunches

1:50–4:50 PM  SBI² Session II: Phenotypic Profiling in Chemical Biology, RNAi Screening and Drug Discovery

Phenotypic profiling represents the development of more complex measures of cellular responses to perturbations in order to identify molecules that affect broad, but well-articulated, changes. Phenotypic profiling and drug discovery aims to identify molecules that influence clinically relevant transitions, including improved glucose management in primary cells, immune function, neuronal dysfunction and tumor development. This session will highlight examples of such systems, including the development of relevant assays and the challenges of advancing hits, including important considerations for counter screens and target deconvolution.

Session Chairs:

Stephanie Mohr (Department of Genetics, Harvard Medical School)

Christopher Moxham (V.P. Quantitative Biology, Eli Lilly & Co)

1:50–2:20 PM  Invited Speaker

High-throughput FISH-based screening approaches to identify regulators of 2D and 3D genome organization

Joyce Eric University of Pennsylvania

Our genome functions as a three-dimensional (3D) chromatin polymer, driven by a complex collection of chromosome interactions. And yet, the mechanisms that control these interactions and drive the spatial organization of the genome remain nearly a complete mystery. To address this gap, we have recently developed two technologies that use fluorescent in situ hybridization (FISH) to interrogate chromosome positioning at single-cell resolution. The first is a technology for high-throughput FISH (Hi-FISH), and the other, called Oligopaints, is a new type of probe that reduces the cost and increases the resolution of FISH. Using these technologies, our work has provided new insights into how chromosomes find and influence each other in the 3D space of the nucleus and has established the Drosophila system of somatic pairing as paradigm for understanding functional genomic organization. I will further describe how these methods should benefit a broad spectrum of research fields, including those focusing on chromosome rearrangements, enhancer-promoter interactions, chromatin compaction, DNA replication, homologous recombination, and copy number variation.

2:20–2:50 PM  Invited Speaker

High content microscopy for cancer drug target and drug safety screens

Danen Erik Division of Toxicology, Cluster Drug and Target Discovery, Leiden Academic Centre for Drug Research, Leiden University

First, an RNAi screening-driven approach for identification of genes involved in cancer progression will be discussed. In this approach, hits regulating cancer cell adhesion/migration are funneled through a pipeline of automated 3D in vitro migration and zebrafish xenograft dissemination assays. Corroborated hits are clinically validated in human cancer specimens and subsequently studied further in vivo using mouse orthotopic transplantation models. Second, the use of a panel of pathway GFP reporter liver cell lines for single cell kinetic analysis of toxicity pathways is discussed. Here, classification of compounds according to activation of distinct stress response pathways is achieved in a time- and concentration-resolved manner by real time microscopy. RNAi screening in this context is used to further unravel adverse outcome pathways.

2:50–3:15 PM  One short talk selected from submitted abstracts

3:15–3:45 PM  Break

3:45–5:00 PM  Three short talks selected from submitted abstracts

5:00–5:30 PM  SBI² Annual General Meeting

5:30–6:30 PM  SBI² Reception

Day 3: Wednesday September 14, 2016  Scientific Program

8:00–9:00 AM  SBI² Poster viewing

9:00–9:10 AM  Day 3 Opening Remarks David Andrews – President SBI²

9:10–12:20 PM  SBI² Session III: High Content Analytics

The power of HCS presents challenges in data analysis. The potential for novel phenotype discovery and functional connections between cellular treatments is great, but this depends on managing the data effectively. Metadata management, data preprocessing and computational analysis are essential steps that need to be integrated into a data analysis process in order to efficiently characterize complex and occasionally subtle changes. This session will present some of the data processing pipeline processes that are essential for analyzing multi-parametric experiments, as well as examples of some of the insights revealed by such studies.

Session Chairs:

Peter Horvath (Biological Research Centre of the Hungarian Academy of Sciences)

Jason Swedlow (Quantitative Cell Biology, University of Dundee, Scotland)

9:10–9:40 AM  Invited Speaker

The Image Data Repository: A Public Resource for High Content Image Data

Rustici G 1 Moore JA 1 Williams E 1 Li S 1 Antal B 3 Chessel A 3 Ferguson R 1 Leo S 1 Sarkans U 2 Tarkowska A 1 Carazo-Salas R 3 Brazma A 2 Swedlow JR 1 1 Division of Computational Biology, University of Dundee, Dundee, United Kingdom 2 EMBL-EBI, Wellcome Trust Genome Campus, Hinxton, United Kingdom 3 University of Cambridge, Cambridge, United Kingdom

Despite major advances in cell and tissue imaging instrumentation and analysis algorithms, key informatics challenges remain unsolved: file formats are proprietary, facilities to store, analyze and query numerical data or analysis results are not routinely available, integration of novel algorithms is hard, and standards for sharing, integrating and publishing data are lacking. To address these challenges, we have built the Image Data Repository (IDR).

The IDR is a prototype platform for publishing, mining and integrating bioimaging data at scale, built to align with requirements published by Euro-BioImaging and ELIXIR and based on the OMERO and Bio-Formats open source software built by the Open Microscopy Environment. It is deployed on an OpenStack cloud running on EMBL-EBI's Embassy resource.

The IDR shows the feasibility and value of publishing image data alongside traditional publications. So far, the IDR includes image data linked to twenty independent studies from genetic, RNAi, chemical, localisation and geographic high content screens, super-resolution microscopy, and digital pathology. Datasets range from several GBs to tens of TBs.

Wherever possible, we sourced and integrated image data with all relevant experimental, imaging and analytic metadata. 89% of the phenotypes recorded by study authors are mapped to ontological annotations, primarily using the CMPO ontology. With this metadata integration, queries of gene names and phenotypes return data from independent but relatable studies. So far, 42 TBs of data and 36 Mio images are available in the IDR at http://idr-demo.openmicroscopy.org.

We have created an open, biological imaging data resource, the IDR, to establish imaging metadata standards and catalyse data integration across a wide range of imaging modalities. In the future, we will add computational facilities to expand the possibilities for data re-use.

9:40–10:10 AM  Invited Speaker

Mining historical ‘microscopy at high-throughput’ screens for novel insights

Ceulemans Hugo 1 Simm Jaak 2 Arany Adam 2 Wegner Jörg 1 Vapirev Alex 1 Chupakhin Vladimir 1 Gustin Emmanuel 1 Singh Shantanu 3 Carpenter Anne 3 Wuyts Roel 2 4 Moreau Yves 2 1 Janssen 2 Katholieke Universiteit Leuven 3 Broad Institute 4 imec

The chemical libraries that pharmaceutical companies screen to select chemical starting points for drug discovery, run into the millions of compounds. At the same time, thousands of internal or external assays are available that can document some kind of biological activity. Even though the number of industrially accessible activity annotations (pxC50 values) of compounds at dose-response quality runs into the 10s-100s of millions, this at first glance impressive number leaves the full compound/assay activity matrix very sparsely populated. Over the last few years, we have been developing a robust machine learning environment to fill in the gaps, integrating multiple single and multi-task classification and regression approaches.

In machine learning exercises to predict compound activity, the conventional input features to describe the compounds are descriptors that encode chemical compound structure. However, we reasoned that any standardized complex readout from an assay that can be acquired at high throughput (250k+compounds) could function as an alternative input space. In particular, recycling historical cellular microscopy screens could make sense. While they have often been designed with a given target or mechanism in mind, a thousand or more potential targets are expressed in the cellular assay and exposed to the full compound library. It is not unreasonable to expect that at least some of these targets induce morphological effects. Therefore, images can be summarized as descriptors that can then be used to try to predict compound activity across a range of assays of interest.

We will describe the results of a first-in-kind exercise on repurposing a historical high throughput microscopy screen covering over half a million compounds for predicting the activities in seemingly unrelated assays.

10:10–10:35 PM  One short talk selected from submitted abstracts

10:35–11:05 AM  Break

11:05–11:55 AM  Two short talks selected from submitted abstracts

12:00–1:30 PM  Lunch Time poster session

1:30–4:00 PM  SBI² Session IV: Innovations in High Content Biology: 3D, Co-Cultures and Stem Cell Systems

A disconnect between easily measured cell systems & in vivo biology is continually recognized as a barrier to a better understanding of developmental biology & disease modification. Image-based screening presents highly sensitive methods to understand complex cellular interactions that are implicit to biology. This session will highlight examples of biologically driven efforts to recapitulate complex biological processes & computational methods needed to quantify changes in these systems.

Session Chairs:

Hunter Elliot (Image and Data Analysis Core, Harvard Medical School)

Michael Mancini (Molecular and Cell Biology, Baylor College of Medicine)

1:30–2:00 PM  Invited Speaker

Next Generation Assays for Next Generation Oncology Targets

Silver Serena Novartis Institutes for BioMedical Research

Growth of cancer cells in 2D format has been a workhorse of the cancer research world, enabling high throughput biology endeavors to identify new targets and new drugs. However, it is clear that we are sampling only a subset of cancer complexity in these models, for example by comparison of genomic characterization between primary tumors, xenografts, and cell lines grown in 2D. I will discuss our efforts to use methods such as co-culture and high content imaging of cells grown in 3D systems to assess if these can indeed “fill the gap” and advance oncology drug discovery.

2:00–2:30 PM  Invited Speaker

High Throughput Analysis of Protein-Protein interactions in cells by Fluorescent Image Moment Analysis

Sergeev Mikhail 1 2 Shah Jagesh V 1 1 Department of Systems Biology, Harvard Medical School and Renal Division, Brigham and Women's Hospital, Boston 2 Université de Montréal and Hôpital Maisonneuve-Rosemont, Montreal QC, Canada

We present the development of a high-throughput microscopy approach to detect protein homo-oligomerization and protein-protein interactions in mammalian cells (fixed or live) using fluorescence image moment analysis. The method takes advantage of the spatial intensity fluctuations to extract oligomerization and/or interaction states of fluorescently labeled proteins. This method has been demonstrated for single confocal images of intact chemically fixed adherent cells. Here we show how this method can be combined with modern confocal high content imaging systems to permit the high-throughput detection of protein-protein interactions in cells.

Using computer simulations and experiments with tandem dimer proteins and small-molecule inducible dimerization domains transiently expressed in human embryonic kidney 293T cells, the detection limits and accuracy of this screening technique will be determined. The developed method is uniquely suitable for large-scale siRNA/CRISPR and small molecule screens where conventional biochemical approaches either fail or require exhausting resources.

2:30–2:55 PM  One short talk selected from submitted abstracts

2:55–3:10 PM  Break

3:10–4:00 PM  Two short talks selected from submitted abstracts

4:00–5:00 PM  Closing Keynote Presentation

High-Throughput Functional Genomics Identifies Regulators of Primary Human Beta Cell Proliferation

Screaton Robert Sunnybrook Research Institute, University of Toronto

The expansion of cells for regenerative therapy will require the genetic dissection of complex regulatory mechanisms governing the proliferation of non-transformed human cells. Discovery-based research using cultured cells, including chemical and functional genomic high throughput screens (HTS), are recognized as powerful, comprehensive tools that identify novel components of signaling pathways. However, as cell lines often display genomic instability and loss of biochemical regulatory mechanisms, screens using primary human cells are now necessary to gain physiological insight in areas most relevant to disease.

The use of primary human cells for HTS introduces several new challenges, including limited source material, cellular quiescence, heterogeneity in cell type, and variability from donor-to-donor. Phenotypic HTS of collections of small molecules can identify lead compounds for cell biological research and hypothesis generation, yet a lack of understanding of the target and the mechanism of action remains a major impediment to their utility due to off target effects.

We have developed a high-throughput RNAi screening strategy specifically for use in primary cells and demonstrate that silencing the cell cycle dependent kinase inhibitors CDKN2C/p18 or CDKN1A/p21 facilitates cell cycle entry of quiescent adult human pancreatic beta cells. This work identifies p18 and p21 as novel targets for promoting proliferation of human beta cells and demonstrates the promise of functional genetic screens for dissecting therapeutically relevant state changes in primary human cells.

Scientific Program and Organizing Committees

Scientific Program Co-Chairs

Steven Haney (Quantitative Biology, Eli Lilly and Company, Indianapolis)

Jennifer Smith (ICCB-Longwood, Harvard Medical School, Boston)

Conference Organization Chair

Doug Bowman (Takeda Pharmaceuticals International Co, Cambridge, Massachusetts)

Educational Program Co-Chairs

Debra Nickischer (Bristol Myers Squibb)

Ann Hoffman (Glaxo Smith Kline)

Marketing

Paul A. Johnston (University of Pittsburgh)