Decoding cis-regulatory element activity through deep learning frameworks

Project Summary Gene expression is tightly regulated by cis-regulatory elements (CREs), which interact with transcription factors and other regulatory proteins to control when and where genes are expressed in different cell types. While millions of CREs have been annotated across the human genome, key questions remain about how these elements work together, how their function varies across cell types, and how they influence alternative isoform usage. Addressing these gaps is essential to understanding the complexity of gene regulation and its role in human health and disease. This project seeks to answer three critical questions: (1) How do cCREs interact in a combinatorial manner to regulate gene transcription? (2) How do cCREs exhibit context-dependent activity in different cell types? (3) How do cCREs contribute to the regulation of transcript isoform usage, beyond simply controlling gene expression? We will answer these questions by leveraging deep learning models, such as convolutional neural networks (CNNs), to analyze large-scale genomic and epigenomic datasets. Deep learning will allow us to uncover non-linear patterns of cCRE interactions and predict their regulatory effects across different cellular contexts. To validate our computational predictions, we will employ high-throughput experimental techniques such as STARR-seq and CRISPR-based perturbation assays. These experiments will assess the regulatory activity of cCREs in their native chromatin environments, allowing for direct testing of combinatorial effects, context-dependent functions, and their impact on transcript isoform expression. Ultimately, this research will provide new insights into the regulatory mechanisms underlying gene expression, with implications for understanding complex diseases and developing targeted therapeutic strategies.