The rapid advancements in high-throughput imaging and sequencing technologies have led to an explosion of omics data in biology, encompassing genomics, transcriptomics, proteomics, metabolomics, and beyond. However, extracting mechanistic insights requires more than just data accumulation; it necessitates integrating and assimilating data arriving from diverse modalities and scales. In this talk, I will discuss a common scenario in omics where statistically independent cross-sectional samples are acquired at a few time points, with the goal of inferring the underlying stochastic process that interpolates the o served high-dimensional marginals. Existing inference approaches based on optimal transport often i pose simplifying assumptions about intrinsic noise, sacrificing accuracy for computational efficiency. We address this challenge by inferring the phase-space probability flow that preserves the same time-dependent marginal distributions as the underlying stochastic process. Our approach, Probability Flow Inference (PFI), disentangles deterministic forces from stochasticity of any form while remaining computationally efficient. In practical applications, we demonstrate that PFI enables accurate parameter and drift estimation in high-dimensional stochastic reaction networks and effectively reconstructs cell differentiation dynamics under substantial molecular noise, surpassing state-of-the-art generative models.
Additionally, a simplified PFI formulation can be leveraged to infer polymer models of chromatin from super-resolution imaging data, providing insights into the interplay between epigenetic (gene expression) and genomic folding (chromatin conformation) landscapes.
Suryanarayana Maddu holds a Bachelor’s degree in Mechanical Engineering from the National Institute of Technology, Karnataka (NITK) and a Master’s degree in Computational Science and Engineering from ETH Zurich and Ruhr University Bochum. He conducted his doctoral research in Computer Science at the Max Planck Institute for Molecular Cell Biology and Genetics in Dresden, Germany. Currently, he is a Flatiron Research Fellow at the Flatiron Institute/Simons Foundation and a visiting scholar at the Q-Bio initiative at Harvard University.