סמינר לפיזיקה של מערכות ביולוגיות וחומרים רכים: Interacting Agents in Biology: Spectral Methods for Cellular Systems and Swarm Configurations

Abstract:

Jonathan Karin's focuses on analyzing biological systems of interacting agents across multiple scales-molecular, cellular, and organismal. Focusing on collective patterns arising from individual interactions. I primarily employ spectral approaches, particularly Graph Signal Processing (GSP), to extract meaningful insights from complex biological data.

In this talk, Jonathan Karin will present two applications at the cellular and organismal levels.

First, he address the challenge of disentangling overlapping signals in single-cell RNA sequencing (scRNA-seq) data, including cell cycle phase, circadian rhythm, and spatial organization. He will introduce scPrisma, which filters and enhances topological signals using spectral template matching, and Annotatability, a framework for identifying annotation mismatches and characterizing biological data structure by monitoring the dynamics and difficulty of training a deep neural network over such annotated data.

Second, he will present his recent work on generating optimal swarm configurations under predation threats using GSP and Generative GNNs. Modeling swarms as dynamic graphs with predation as graph signals revealed a detectability-durability trade-off: swarms must balance avoiding detection with surviving once detected. Leveraging this insight, we built a generative model that produces resilient swarm configurations.