Developing a cCRE-Centric Infrastructure in AnVIL to Characterize Variant Effects on Gene Regulation

About This Grant

Project Summary Understanding how genetic variation impacts gene regulation is essential for linking noncoding variants to disease risk and transcriptional dysregulation. While cis-regulatory elements (cCREs), such as enhancers and promoters, play a central role in transcriptional control, their activity is highly cell type-specific, and most allele specific regulatory studies have been conducted at the bulk tissue level, limiting resolution. Additionally, previous studies have largely focused on gene-level expression changes, overlooking how regulatory variation affects alternative isoform usage, which has important implications for human disease. This project will integrate large-scale regulatory annotations with allele-specific analyses at the single-cell level to improve our understanding of how noncoding variation shapes transcriptional regulation. Aim 1 will establish a cCRE workspace and analysis framework in AnVIL, integrating the ENCODE Registry of cCREs into a cloud-based platform to support scalable and reproducible analyses of transcriptional regulation. We will develop modular workflows for cCRE annotation, cell type-specific scoring, and transcription factor footprinting, along with the STELLA suite, a collection of computational tools for regulatory genomics. Aim 2 will investigate how allele-specific cCRE activity influences isoform usage in individual cell types using data from the Genomic Answers for Kids (GA4K) project. We will construct personalized diploid genomes to identify allele-specific cCREs (from single-cell ATAC-seq) and allele-specific isoform usage (from bulk long-read RNA-seq). Using a Dirichlet-Multinomial deconvolution model, we will infer cell type-specific isoform expression, validated with ENCODE Split-seq data. We will then use a hierarchical regression model to test whether allele-specific cCREs predict allele-specific isoform usage, incorporating cell type assignment probabilities. Finally, we will apply ChromBPNet deep learning models to assess the functional impact of noncoding variants on transcription factor binding. By integrating these analyses into AnVIL, this project will create generalizable computational frameworks for regulatory genomics, enhancing the usability of NHGRI-funded resources. The methods and resources developed will enable broad applications across diverse datasets and disease studies, ultimately improving our ability to interpret noncoding variation in gene regulation and human disease.