QuBBD: Mathematical models for a molecular genetic understanding of population variation in risk of cardiovascular disease

Oregon State University, Corvallis OR

Investigators

Abstract

Genome-wide association studies (GWAS) are a powerful approach for mapping the regions of the genome containing single-nucleotide differences in the population (single-nucleotide variants or SNVs) that are associated with trait variation. Narrowing down from GWAS-identified genomic regions to the individual SNVs that are responsible for trait variation is particularly challenging for regions of the genome that are in-between genes. A second challenge in GWAS is that the large number of SNVs necessitates a high level of statistical stringency, and thus, many biologically relevant SNVs are missed. GWAS is used for many human disease traits (such as coronary artery disease or CAD), and thus, addressing these two challenges would have broad significance in biology and biomedical research. This award supports initiation of a collaborative research project that will address these two challenges by developing mathematical models that integrate a variety of types of measurements and information derived from cells and population studies, in order to pinpoint SNVs in-between genes that affect trait variation, and to improve the statistical power of GWAS. CAD is a high-significance application for improving GWAS because of CAD's prevalence (15 million in the U.S.). The objectives of this project are to (1) create and evaluate an integrative statistical model for improving power for GWAS analysis and for discovering novel gene-trait associations and (2) create and evaluate a machine-learning model for identifying regulatory variants within intergenic GWAS regions. The models would incorporate features from large-scale datasets from the Framingham Heart Study SHARe database, the ENCODE project, the GTEx project, and CARDIOGRAMplusC4D. The models' performance would be benchmarked against previously published models. The project's significant outcomes would be: (1) the first analytic statistical model for integrative GWAS analysis that would provide a readily interpretable significance score; (2) a quantitatively validated and interpretable machine-learning model for combining genomic information types to predict regulatory variants; (3) feature importance scores for the genomic features that are integrated within the model; and (4) identification of new GWAS loci for population variation in CAD risk, and, for regulatory variants within the loci, the genes and transcription factors (and ultimately, the gene functional annotations) with which they are associated. Software implementations of the methods will be shared in an open-source software repository. This award is supported by the National Institutes of Health Big Data to Knowledge (BD2K) Initiative in partnership with the National Science Foundation Division of Mathematical Sciences.