Developing novel probabilistic models to… (F31HG015202) – Project Grant to Regents Of The University Of…

Abstract: Genome-wide association studies (gwas) have uncovered thousands of associations between genetic variants and clinically relevant diseases and traits. These associations provide information about clinical, environmental, and molecular causes of disease. Mendelian randomization (MR) is a statistical method that can be used to estimate a causal effect between a measured exposure and outcome trait using genetic associations with exposure and outcome traits. For accurate MR estimation, exposure and outcome genetic studies must be sampled from the same population. However, this limits the power of MR to detect causal effects in populations that do not have large gwas, and prevents researchers for making use of large, highly-powered meta analyses. In aim 1, i will develop a novel MR estimation method, metamr, which can leverage multiple gwas studies for the exposure trait to improve precision and accuracy of the causal estimate. Metamr will produce more accurate causal effect estimates and have wider applicability than existing MR estimation methods. Gwas associations also contain information about unmeasured molecular pathways contributing to disease that cannot be evaluated using MR. Previous studies have shown that genetic associations across sets of related traits can be decomposed into common patterns, or factors, that reflect broad biological functions. However, without additional information, IT is difficult to connect these factors to specific molecular pathways, limiting their clinical utility. Variant annotations are sets of auxiliary information about variant function such as chromatin accessibility, associations with gene expression, and kegg pathway annotations associated with the nearest gene. Past work has shown that variant annotations can be used to identify disease relevant tissues and molecular functions via enrichment analysis, and that incorporating annotations can improve power to identify disease-associated variants. In aim 2 i propose an annotation-informed cross-phenotype factorization method. This tool will identify groups of variants with shared phenotypic effects and similar annotations, improving power to identify factors, and providing natural biological interpretations. I will apply the method developed in aim 2 to understand molecular contributors to cardiovascular disease, using clinical traits associated with cardiovascular disease and annotations including protein quantitative trait loci identified in topmed and kegg pathway gene annotations. By completing these aims, i will gain expertise in developing statistical methods for genetic data, as well as scientific writing and presentation. I will grow as a scientific collaborator by integrating myself into the topmed community as part of the application of aim 2. In addition to research activities, i will further my teaching skills through classroom instruction on teaching and giving guest lectures and seminars, and deepen my connections with the wider scientific community by attending conferences. Overall, completing the activities proposed under this fellowship will allow me to hone the skills i need to become an independent researcher.

Developing novel probabilistic models to identify causal contributors to disease

Award facts

Description

Quick answers

Who received award F31HG015202?

How much is award F31HG015202 worth?

See this award's full record