Modeling and Predicting Subcellular Transcript Organization in Situ

XIANG ZHOU – YALE UNIVERSITY 

ABSTRACT

Subcellular spatial transcriptomics provides an unprecedented view of how RNA molecules are organized within individual cells, offering new opportunities to understand cellular function in their native tissue context. In this talk, I will present two complementary approaches for modeling and interpreting subcellular transcript organization. First, I will introduce ELLA, a statistical framework for detecting and characterizing spatial variation in RNA localization within cells, and briefly discuss its extension to two-dimensional cellular representations. I will then present SVC, a Vision Transformer-based framework that integrates subcellular transcript localization with gene function, cell morphology, cell type, and tissue microenvironment to build spatially grounded representations of genes and cells. SVC enables prediction of subcellular expression patterns for unmeasured genes, spatial imputation, characterization of gene localization similarity, and in silico modeling of cellular perturbations. Together, these approaches illustrate how statistical and AI models can move us from describing where RNA is within cells to predicting and modeling cellular organization in situ.