Statistics Seminar: Spatially Varying Gene Regulatory Networks via Bayesian Nonparametric Covariate-Dependent Directed Cyclic Graphical Models
Presenter: Trisha Dawn, UC Santa Cruz
Description: Spatial transcriptomics technologies enable measurement of gene expression with spatial context, providing opportunities to understand how gene regulatory networks (GRNs) vary across tissue regions. Studying spatial variation in GRNs is crucial for uncovering mechanisms governing heterogeneous cellular functions and gives rise to a challenging statistical problem: graph learning with spatially heterogeneous data. Existing methods focus on undirected graphs or directed acyclic graphs, limiting their ability to capture feedback loops in gene regulation. Although cyclic graphs have been used to construct GRNs, ensuring their stability (an eigenvalue constraint) remains statistically and computationally challenging, especially while allowing graph structure to vary spatially. We propose BNP-DCGx, a Bayesian nonparametric covariate-dependent directed cyclic graphical model for learning spatially varying gene regulatory networks (svGRNs). The key idea is to introduce a covariate-dependent random partition that discretizes the covariate space into clusters, each with a cluster-specific stable directed cyclic graph. Through partition averaging, we obtain smoothly varying graph structures over space while preserving stability and providing large prior support for continuously varying graphs.
About the speaker: My research centers on the development of novel statistical methods mostly in causal discovery, change-point detection and graph matching, drawing on techniques from Bayesian nonparametrics, optimization, machine learning, and high-dimensional inference. I am particularly interested in applying statistical methodology to emerging data-rich fields such as spatial transcriptomics, single-cell multi-omics, network analysis, and dynamically evolving complex systems. Currently I am developing statistical methodologies to detect changes in microbial communities using cytogram data collected across varying environmental conditions and research cruise trajectories at Dr. Hyun’s group as a postdoc. I earned Ph.D. in Statistics from Texas A&M University, jointly advised by Yang Ni and Jesús Arroyo. Prior to my doctoral studies, I completed Master’s degree at the Indian Statistical Institute and joined as a Junior Research Fellow. I also obtained Bachelor’s degree in Statistics at St. Xavier’s College (Autonomous), Kolkata.
This seminar is hosted by Professor Yunyi Shen