Biological & Soft Matter Seminar: Inference of signal related properties in single cell data using topological assumptions

Abstarct:

Single-cell RNA sequencing (scRNA-seq) generates detailed gene expression profiles that reveal complex cellular states, including cell type, cell cycle phase, gene regulatory patterns, and tissue localization. However, disentangling these signals remains a significant challenge. In this talk, I will present two recent approaches to address this task: scPrisma, which filters and enhances topological signals in the data using spectral template matching, and Annotatability, which interprets training dynamics in deep learning models to improve the analysis of single-cell data. The talk will highlight biological insights gained from applying these methods, such as identifying circadian rhythm-related regulatory networks and inferring intermediate cell states along the epithelial-mesenchymal transition, while also discussing the challenges and limitations of these approaches.