PQG Seminar

Tao Wang
Associate Professor, Department of Bioinformatics and Computational Biology, University of Texas MD Anderson Cancer Center

Interpretable AI Model Reveals Spatial Codes of Cell Recruitment in Tissues

Understanding which molecular signals recruit specific cells to precise tissue locations is fundamental to tissue biology, immunity, and disease. Existing computational tools for spatially resolved transcriptomics (SRT) describe where cells reside but cannot explain why. Here we introduce SPACER, a fully interpretable multi-instance learning framework that decodes gene-level recruitment rules directly from SRT data. Applied to 37 SRT datasets spanning multiple human tumor types and mouse heart, SPACER achieved 4-21 fold improvement over existing methods in various tasks of recovering biologically meaningful recruitment signals. SPACER revealed (1) that tumor-expressed mucins are associated with physical barriers to T cells engagement, which predicts resistance to immune checkpoint inhibitors (ICIs) and patient prognosis, (2) that local antigen presentation is the dominant correlate of T cell engagement across cancer types, and (3) that Cd4⁺ T cells are the dominant functional effectors in severe ICI-induced myocarditis. These findings establish SPACER as a discovery engine for cellular recruitment biology with direct translational applications.

Shulei Wang
Associate Professor, Statistics
Associate Professor, Nutritional Sciences
University of Illinois Urbana-Champaign

Making Sense of Whole-Cell Models: COTree for Cell-Resolved Multi-Omics Trajectories

The ambition of building virtual cells raises a fundamental question: how can computational models help us understand cellular behavior? Mechanistic whole-cell models offer one route by simulating molecular processes and their interactions over time. Yet specifying reaction mechanisms does not directly reveal the collective growth patterns, alternative outcomes, or early signs of failure that emerge across simulated cells.

In this talk, I will introduce COTree, a statistical framework for interpreting cell-resolved multi-omics trajectories from whole-cell models. COTree first learns a compact representation through complementary reconstruction and prediction tasks, capturing information about both current molecular states and their temporal evolution. It then constructs a time-informed trajectory principal tree that summarizes shared progression paths and branching outcomes while preserving within-cell temporal order. These representations support cell classification, early fate prediction, and the identification of molecular changes associated with trajectory divergence.

Using simulations of the minimal cell JCVI-syn3A, I will illustrate how COTree distinguishes replication outcomes and identifies early molecular signatures of glycolytic failure before simulated cell death. The talk highlights the complementary roles of mechanistic modeling and statistical learning in turning complex simulation output into interpretable patterns and testable biological hypotheses.

Graham McVicker
Associate Professor
Laboratory of Genetics, Salk Institute

Ritambhara Singh
Associate Professor of Computer Science at the Siebel School of Computing and Data Science (SSCDS), University of Illinois Urbana-Champaign (UIUC)

Jingshen Wang
Associate Professor, Biostatistics
UC Berkeley, Public Health

Gal Mishne
Associate Professor, Halıcıoğlu Data Science Institute (HDSI)
University of California, San Diego