Spatial Organization of Biological Fuctions | BPS Thematic Meeting

Spatial Organization of Biological Functions Meeting

Thursday Speaker Abstracts

GEOMETRY-AWARE MULTI-OMICS INTEGRATION: UNIFYING GAWRDENMAP AND PROXIMOGRAMS TO DECIPHER TISSUE SPATIAL ORGANIZATION Arvind Rao ; 1 University of Michigan Ann Arbor, Computational Medicine and Bioinformatics, Ann Arbor, MI, USA Deciphering the spatial organization of biological functions within tissues necessitates frameworks that can seamlessly integrate multi-modal data while preserving the intrinsic geometries of cellular interactions. In this presentation, we introduce a unified approach that amalgamates two complementary computational frameworks—GardenMap and Proximogram— to analyze tissue architecture through the lenses of spatial statistics and Riemannian geometry. GaWRDenMap employs geographically weighted regression (GWR) to quantify spatially varying interactions between cell types, such as epithelial and immune cells, across tissue sections. By transforming the distributions of GWR coefficients into probability density functions (PDFs) and mapping them onto a Riemannian manifold of square-root densities, we capture the nuanced variations in cellular interactions. This geometric representation facilitates the computation of intrinsic distances and supports principal component analysis within the manifold, enabling robust classification of tissue states based on spatial interaction signatures. Complementing this, Proximogram constructs graph-based representations that integrate spatial imaging data with single-cell omics profiles. By embedding independently acquired datasets into a joint graph structure, Proximogram captures both molecular profiles and spatial contexts. Utilizing graph convolutional networks (GCNs), this framework enhances the classification of disease states and uncovers spatially informed biomarkers, demonstrating improved discriminatory power over models relying solely on spatial data. Together, these frameworks provide a comprehensive toolkit for analyzing the spatial organization of biological functions. By integrating spatial statistics, Riemannian geometry, and graph-based multi-omics analysis, we offer novel insights into tissue heterogeneity and disease pathology. This unified approach holds promise for advancing our understanding of complex biological systems and informing therapeutic strategies.

TBD Maithreyi Narasimha Tata Institute of Fundamental Research, India No Abstract

24