From learning to leveraging graphs in bioinformatics
Weekly Wednesday Wartik Genomics Lecture Series

Graphs are a natural language for biology: capturing relationships between genes, cells, and tissues. In this talk, I will present two complementary perspectives on how graphs can be both learned from and leveraged in bioinformatics.

First, I will introduce OTVelo, a method for inferring gene regulatory networks from time-stamped single-cell gene expression data. OTVelo estimates gene velocities using optimal transport when sequencing depth limits direct measurement, then infers dynamic regulatory relationships across time points via time-lagged correlation and Granger causality. Rather than collapsing regulation into a single static network, OTVelo uncovers the underlying temporal mechanisms driving gene regulation.

Second, I will present DRIFT, a framework that leverages spatial graphs to enrich foundation model representations for spatial transcriptomics (ST). While powerful single-cell foundation models have been trained on large-scale scRNA-seq data, they do not exploit the spatial structure inherent to ST data. DRIFT addresses this gap by propagating cell embeddings across spatial neighborhoods using heat kernel diffusion, incorporating local context without retraining existing models.

Together, these works illustrate a broader theme: graphs are not merely a representation choice, but a powerful inductive bias that enables richer biological discovery.

Bio: Ritambhara Singh is an Associate Professor of Computer Science and Data Science and a member of the Center for Computational Molecular Biology at Brown University. Her research lab develops machine learning methods with the goals of data integration and model interpretation for biological and biomedical applications. Prior to joining Brown, she was a post-doctoral researcher in the Noble Lab at the University of Washington. She completed her Ph.D. in 2018 from the University of Virginia with Dr. Yanjun Qi as her advisor. Ritambhara has received the NHGRI Genomic Innovator Award and Brown University’s Richard B. Salomon Faculty Research Award for developing deep learning methods to integrate and model genomics datasets. She has also received the Dean’s Award for Excellence in Teaching at Brown. She recently received the NSF CAREER award for developing integrative and explainable machine learning methods for heterogeneous health-related datasets.

Contact

  Donna McMinn
  dlp18@psu.edu
  +1 814-935-3444