Title: Reconstructing Transcriptional Regulatory Mechanisms from Multi-Omics Data

Abstract: Reconstruction of gene regulatory networks (GRNs), specifically those connecting transcription factors (TFs) to their target genes, has been a major challenge for decades. Such networks can help us identify major TFs and TF-gene relationships that underlie expression changes in the biological process of interest, such as development or disease progression. Frequently, GRN reconstruction methods exploit correlations in expression data and in many cases they also utilize genome-wide TF-DNA binding and epigenomic profiles. Recent work in my group has sought to address the GRN reconstruction problem while integrating multiple types of data pertaining to TF-gene relationships. I will present the main underlying methodology, based on probabilistic models, and its application to two biological problems: (i) identifying TFs underlying drug response variation among individuals, while integrating genotype, TF-DNA binding, DNA methylation, gene expression and cytotoxicity data, and (ii) identifying TFs underlying colorectal cancer progression from dynamic histone modification data as well as gene expression and TF-DNA binding data.

In the last segment of the talk, I will present our on-going work on deciphering how enhancers sequences encode regulatory function. We adapted a thermodynamics-based model of the sequence-to-expression relationship to study the transcriptional program driven by the TF Estrogen Receptor a in response to estrogen treatment in breast cancer cells. The model not only helps us reconstruct the associated GRN, it also allows us to mechanistically interpret single-nucleotide variants associated with expression and phenotype, thus providing a potential computational approach to the genotype-to-phenotype problem