Abstract: Recent spatial transcriptomics (ST) technologies make high-throughput measurements of gene expression at thousands of locations in a 2-D tissue slice. However, due to current ST technological limitations, these measurements are highly sparse—thus complicating the identification and analysis of spatial gene expression patterns. In this talk, I will present computational and machine learning approaches that overcome these technological limitations by modeling the latent geometry of a 2-D tissue slice. First, I will present Belayer, an algorithm for learning both discrete and continuous variation in layered tissues using tools from complex analysis (conformal mapping, harmonic functions). Second, I will present GASTON and GASTON-Mix, unsupervised and interpretable deep learning algorithms to learn a “topographic map” of a 2-D tissue slice. I will show how our algorithms uncover subtle gene expression patterns across a diverse range of biological systems including the brain, skin, liver, and tumor microenvironment.
Bio: Uthsav Chitra is an Assistant Professor of Computer Science and faculty in the Data Science and AI Institute at the Johns Hopkins University. His research broadly develops statistical/machine learning methods for analyzing high-dimensional and multi-modal biological data, with a particular focus on spatial and graph-based models. His research has been recognized with a Rising Stars in Data Science award, a Best Paper Award at RECOMB-CCB, a Siebel Scholar award, and an NSF Graduate Research Fellowship. Uthsav holds a PhD in Computer Science from Princeton University (2024), and an ScB in Mathematics, an AB in Computer Science, and an AB in Applied Math from Brown University in 2017. Uthsav was previously a postdoctoral fellow at the Eric and Wendy Schmidt Center at the Broad Institute of MIT and Harvard (2024-2025) and a software engineer at Facebook (2017-2018).