High-throughput computational modeling to assess the role of 3D genome folding in human congenital anomalies

PROJECT SUMMARY Congenital heart defects (CHD) occur in nearly one percent of live births each year and are the leading cause of defect-associated infant mortality. The majority of genetic studies in CHD have focused on variation within the protein-coding exome; however, most disease-risk loci fall in noncoding regions, and it is presumed that

some of these represent important regulators of gene expression such as cis-acting enhancers and insulators. In spite of these studies, less than half of the heritability of CHD has been explained via genome-wide associations or burden testing of protein-coding genes and putative regulatory elements. In this proposal, we

hypothesize that genetic variants that alter 3D genome folding contribute to the etiology of CHD by disrupting the contacts of key cis-regulatory mechanisms in development. One type of chromatin structure that could be affected is the Topologically Associating Domain (TAD), which refers to a level of chromatin organization

characterized by higher contact frequency within the domain relative to loci outside of that domain. It has been shown that while unaffected controls show a clear depletion of SVs at TAD boundary regions across the genome, individuals diagnosed with Developmental Delay (DD) and autism showed no bias in the genomic

location of SVs. Based on these findings and the high rates of co-morbidity between DD and CHD, the first aim characterizes structural variation at TAD boundaries and other non-coding regulatory regions in CHD relative to controls, and will further determine whether TADs are enriched for SVs in a region-specific manner based on

proximity to genes active in the developing heart. In the second aim, we use a complementary annotation- agnostic deep learning approach developed in our group to predict chromatin contact changes as a result of genetic variants in CHD patients. We will use Hi-C sequencing to confirm the model-predicted chromatin

contact effects in iPSC-derived endothelial cells and cardiomyocytes, and additionally use RNA sequencing to determine whether the hypothesized transcriptional disruption occurred. Finally, in the third aim we will create a model that uses genetic, epigenetic, and transcriptional features of genetic variants found in each individual

to predict their phenotypic status. By analyzing the relative importance of features used to make predictions, we will be able to determine what types of biological mechanisms and pathways are most predictive for congenital heart anomalies. Collectively, these findings will elucidate a currently underappreciated source of

regulatory disruption in human development, identify new disease-relevant genes and potential therapeutic targets, and will refine and validate a method for the high-throughput prediction of chromatin contact frequency to advance the field of 3D genomics.