EWSC - From Model Explanations to Discovery: Explainable AI in Cancer Precision Medicine (2026) @broadinstitute

Uploaded May 2026 | Updated September 2026, 2 weeks ago

EWSC - MIT EECS Colloquium
February 3, 2026
Broad Institute of MIT and Harvard

Su-In Lee
Boeing Endowed Professor, Paul G. Allen School of Computer Science & Engineering, University of Washington

Abstract:
Explainable AI (XAI) has made significant strides in recent years, offering valuable theories and techniques to interpret complex machine learning models. However, these methods often struggle when applied to interpreting complex datasets for scientific discovery, particularly those involving high-dimensional omics data such as gene expression profiles. These datasets, crucial for understanding biological systems, require novel approaches to fully unlock the potential of XAI. In this talk, I will delve into the practical challenges of applying XAI to gene expression data, showcasing case studies that underscore its potential and limitations. I will present innovative strategies for adapting XAI techniques to accelerate data-driven discoveries in cancer pharmacology and cancer systems biology. The discussion will illuminate how addressing these challenges can lead to profound biological insights and impactful clinical implications. By bridging the gap between advanced XAI principles and techniques and the demands of real-world biomedical datasets, this talk aims to inspire the development of more robust methodologies at the intersection of AI and biomedicine, paving the way for a new era of innovation in biomedical research.

Biography:
Professor Su-In Lee is the Boeing Endowed Professor of Computer Science at the University of Washington (UW). She earned her Ph.D. from Stanford University in 2009 under Professor Daphne Koller and joined UW in 2010 after serving as a Visiting Assistant Professor at Carnegie Mellon University. She is renowned for her groundbreaking research at the intersection of AI, biology, and medicine, and widely recognized as a pioneer in explainable AI (XAI). Her seminal contributions include the SHAP framework and its subsequent algorithms and principles, which have fundamentally transformed the interpretation of machine learning models across disciplines. She has been honored with major awards, including the Samsung Ho-Am Prize in Engineering (the “Korean Nobel Prize,” as its first woman recipient in 34 years), the ISCB Innovator Award, and the NSF CAREER Award. She is an American Cancer Society Research Scholar, an AIMBE Fellow, and an ISCB Distinguished Fellow. Her recent work advances fundamental principles of XAI and applies them to biomedicine—from uncovering molecular drivers of disease to auditing clinical AI systems—fundamentally reshaping how AI is integrated into biomedical research and healthcare. This integration has enabled novel discoveries and produced numerous awards and highly cited publications spanning AI, molecular biology, and clinical medicine.

For more information, visit: broadinstitute.org

Copyright Broad Institute, 2026. All rights reserved.

EWSC - From Model Explanations to Discovery: Explainable AI in Cancer Precision Medicine (2026)

$MIA: Lorin Crawford, Rethinking scale in single-cell foundation models; Primer: Davide DAscenzo Models, Inference and Algorithms September 24, 2025 Broad Institute of MIT and Harvard Meeting: When more isn’t better: rethinking scale in single-cell foundation models Lorin Crawford Microsoft Research New England (MSR) The success of transformer-based foundation models on natural language and images has motivated their use in single-cell biology. Single-cell foundation models have been trained on increasingly larger transcriptomic datasets, scaling from initial studies with 1 million cells to newer atlases with over 100 million cells. In this talk, we will investigate the role of pre-training dataset size and diversity on the performance of single-cell foundation models on both zero-shot and fine-tuned tasks. In the first half, we use a large corpus of 22.2 million cells to pre-train a total of 400 models and evaluate over 6,400 experiments. We show that current methods tend to plateau in performance with pre-training datasets that are only a fraction of the size. This will lead us to the second half of the talk where we evaluate training data composition on model performance. Focusing on a tractable biological system (human hematopoiesis), we train and analyze deep generative models with a variety of training datasets, including cells from adult and developing tissues, disease states, and perturbation atlases. From the performance across these models, we observe that (1) deep generative models generalize poorly to unseen cell types, (2) addition of malignant cells to a healthy cell training corpus does not necessarily improve modeling of unseen malignant cells, and (3) inclusion of an embryonic stem cell transcription factor differentiation atlas in training data improves performance on out-of-distribution tasks. Our results highlight the distinct contributions of different training data types and point towards strategies for optimizing future single cell foundation models. Primer: Infrastructure and modeling challenges in single-cell omics Davide DAscenzo Polytechnic University of Turin Deep learning is expected to play a major role in advancing the analysis of single-cell omics data, but a number of practical challenges have so far limited its impact. In this talk, we will focus on two such challenges, one at the level of data infrastructure and one at the level of model design. With the rise of very large-scale single-cell experiments, we are now able to use datasets of hundreds of millions of cells. Model training on these datasets is often bottlenecked by the data-loading process of moving data from disk to GPU. To address this, we developed scDataset, a scalable and efficient data-loading solution for quasi-random sampling of single-cell data from disk [1]. We present block sampling and batched fetching strategies that balance I/O efficiency with memory consumption and minibatch diversity, demonstrating how to achieve increased throughput while maintaining sampling quality comparable to true random shuffling. Moving to biological tasks, we focus on cell type annotation. We find that both simple linear models and more complex transformer-based architectures struggle to generalize in out-of-distribution settings. To mitigate this, we introduce a hierarchical cross-entropy loss that incorporates the structure of cell type ontologies [2]. Across model classes, this leads to consistent improvements in performance, suggesting that structured biological priors may be more useful than scaling up model parameters.[1] Davide D’Ascenzo and Sebastiano Cultrera di Montesano. scDataset: Scalable data loading for deep learning on large-scale single-cell omics. arXiv, 2025. [2] Sebastiano Cultrera di Montesano, Davide D’Ascenzo, Srivatsan Raghavan, Ava P. Amini, Peter S. Winter, and Lorin Crawford. Hierarchical cross-entropy loss improves atlas-scale single-cell annotation models. bioRxiv, 2025. For more information visit: https://www.broadinstitute.org/talks/spring-2025/mia Copyright Broad Institute, 2025. All rights reserved.$

MPG Primer: Beyond the Gene: A Not So Deep Dive into Bulk RNA-Seq, Isoforms & Alternative Splicing

ML4H: Collin Stultz: Self-Supervised Learning for Patient Phenotyping and Forecasting Adverse Events

MPG Primer: The Awesome Power of Genetic Linkage (2025)

MIA: Faisal Mahmood; Multimodal, Generative, and Agentic AI for Pathology

MPG Primer: Gene-environment interactions: a genome-wide perspective (2025)

Meet Alyssa Lawler - Postdoctoral Fellow #Science #MentalHealth

Variant to Function (V2F) Symposium: Keynote, Robert Plenge

Cancer-Specific Foundation Models: Friend or Foe in Healthcare AI?