| Speaker | Topic | Abstract |
|---|---|---|
| Shantanu Singh Broad Institute |
Morphology at scale for functional genomics and drug discovery | TBD |
| Christina S Leslie Memorial Sloan Kettering Cancer Center |
Discrete representation learning for imaging-based spatial transcriptomics | Recent imaging-based spatial transcriptomics (imST) platforms enable the high-resolution localization of transcripts for 500-5000 genes in intact tissue slices, providing unprecedented spatially resolved gene expression data to dissect cellular ecosystems. Most computational pipelines for imST start with cell segmentation, assigning transcripts to cells based on inferred cell boundaries to yield a cell x gene matrix. This allows the use of standard single-cell analysis methods but incurs many errors and filters out low-count cells. We describe an alternative strategy based on discrete representation learning, yielding a codebook representing multi-scale neighborhoods as well as cell types. This strategy enables a more complete cell annotation and more reliable downstream analyses. |
| Juan C Caicedo Morgridge Institute for Research |
Toward foundation models of cellular morphology. |
Cellular morphology is a biologically meaningful readout that can be obtained from microscopy images of different types, and has applications in drug discovery and functional genomics. Deep learning models for cellular morphology are often trained for a specific type of microscopy image, such as brightfield or a fluorescent panel with a set number of channels. This makes reusing models from one experiment to another difficult, because the channels do not always match across experiments. Here, we describe our efforts to create foundation models that can be reused across different types of imaging experiments, which are channel adaptive, don’t require training from scratch, and can match or surpass the performance of specialized models. |
| Jean Fan Johns Hopkins Biomedical Engineering |
Multi-sample comparative spatial omics data analysis |
Recent advances in high-throughput spatially resolved transcriptomics technologies now enable high-throughput molecular profiling of cells while maintaining their spatial organization within tissues. Application of these technologies provides the opportunity to contribute to a more complete understanding of how cellular spatial organization relates to tissue function and how cellular spatial organization is altered in disease. New statistical approaches and scalable computational tools are needed to connect these molecular states and spatial-contextual differences. In this talk, I will provide an overview of spatially resolved transcriptomics technologies and associated computational analysis methods developed by my lab. Specifically, to facilitate spatial molecular comparisons across structurally matched tissue sections from replicates, we developed STalign to align ST datasets in a manner that accounts for partially matched tissue sections and other local non-linear distortions using diffeomorphic metric mapping. Likewise, to facilitate comparison cell-type spatial organizational patterns, we developed CRAWDAD, Cell-type Relationship Analysis Workflow Done Across Distances, to quantify pair-wise cell-type spatial relationships across length scales. We demonstrate how such multi-scale characterization enabled by CRAWDAD can be used to compare cell-type spatial relationships across multiple samples to identify consistent as well as patient and sample-specific cell-type spatial relationships. We anticipate that such statistical approaches and computational methods for analyzing spatially resolved transcriptomic data will offer the potential to identify and characterize the heterogeneity of cells within their spatial contexts and contribute to important fundamental biological insights regarding how tissues are organized in both the healthy and diseased settings. |
| Patrick Schwab GSK |
TBD | |
| Peter Horvath Szeged, Helsinki, ETH Zurich, INRIA, Sophia Antipolis |
Life beyond the pixels - single-cell analysis | |
| Daniel Racoceanu Sorbonne University |
Explainable Artificial Intelligence (XAI): a modern Ariadne’s thread in Biomedical Imaging. Supporting Discovery of new knowledge, Frugal environmental respecflul Computational Approaches. | Two use cases : PhagoStat - an efficient quantification of cell phagocytosis in neurodegenerative disease studies / Virtual staining - Scalable, Trustworthy Generative Model for Virtual Multi-Staining from H&E Whole Slide Images. PhagoStat a scalable and interpretable end to end framework for efficient quantification of cell phagocytosis in neurodegenerative disease studies : This study presents a real-time, scalable, and interpretable deep learning pipeline for quantifying phagocytosis in dynamic, unstained cells from phase-contrast video microscopy—an important task for neurodegenerative disease research. The framework handles large datasets, includes quality control to address issues like blurring or movement, and features explainable segmentation to improve transparency over traditional black-box models. The pipeline balances interpretability and high performance, optimizing algorithm design and execution speed. Applied to microglial cells in frontotemporal dementia (FTD), it reveals that FTD mutant cells are larger and more phagocytic than controls. We released our method and dataset as open-source tools to support future research in neurodegeneration. |
| Gayathri Mohankumar AstraZeneca |
AI Science in Drug Discovery and Development | |
| Nicolas Brieu AstraZeneca |
AI Science in Drug Discovery and Development |