Perturbation-response approaches to determining the regulatory networks underlying human complex traits

Project Summary/Abstract A large majority of heritable human traits and diseases are complex, with genetic variants of small effect spread throughout the genome. It is now understood that genetic contributions to disease are enriched in gene promoters and enhancers thought to regulate expression. This has led to the hypothesis that genetic variation leads to disease via disruption of an underlying gene regulatory network, either via trait-relevant pathways or distal perturbations propagating through the network to trait-specific core genes. At present, our lack of understanding of the networks themselves limits our ability to understand how their disruption can lead to disease state. In the long-term, causal models of these networks may reveal avenues for treatment by suggesting mechanisms for returning the system to proper functioning. Here, I propose to leverage recent developments in causal inference to show that novel computational methods enable integration across large-scale data generation efforts to highlight regulatory changes underlying common disease. I propose to i) improve causal structure learning methods to better leverage prior biological knowledge and improve network estimation for genes that are lowly expressed, poorly captured or difficult to intervene on experimentally and ii) construct a causal network integrating populationscale eQTL data and GWAS summary statistics, and conduct a thorough comparative analysis with large-scale CRISPR perturbation data in immortalized cell lines. The first aim will enable us to construct genome-wide causal networks using single cell CRISPR screen data, including many functionally-relevant genes that are difficult to capture using existing methods. Our second aim will enable identification of core disease-relevant genes and their pathways, and allow us to identify which traits and disease pathways are best studied by current in vitro immortalized cell line models. Completion of these aims will provide a framework for large-scale estimation of regulatory networks and their role in complex trait biology.