Project Summary Polygenic scores (PGS), constructed by common variants identified through large genome-wide association studies (GWAS), are effective tools in research and clinical applications. However, the interpretation of the functional roles of PGS—that involve hundreds to even thousands of common variants—in uncovering the specific components of a phenotype trait or a disease outcome is challenging. This will further obscure the interpretations of downstream analysis in identifying the causal relationships between risk factors and diseases. The proposed project aims to address a critical need to enhance the PGS applicability in investigating disease etiologies using cohort and family-based studies, integrating diverse sources of information including multi-omics data and environmental exposures. In particular, the work will 1) Develop a machine learning approach and its hypothesis testing framework to integrate trait-associated SNPs, multi-omics data, and summary-level statistics of the trait to model the mediating mechanisms underlying genetic associations. Then construct partitioned SNP sets or partitioned PGS (pPGS), with each partitioned component representing distinct functional regulatory pathways linked to the GWAS trait. 2) Develop likelihood-based methods utilizing multi-trait PGS to estimate the causal effects of multiple correlated exposures on an index disease in family-based studies, correcting for biases from assortative mating and population stratification. 3) Identify heterogeneous causal effects of exposures using partitioned SNP sets and pPGSxE interactions in leading causes of mortality, including cardiovascular diseases and cancers; and identify the causal relationships of multiple maternally-mediated exposures and biomarkers on childhood diseases, including autism spectrum disorders and orofacial clefts, through multi-ethnic case-parent trio studies. Finally, the developed methods and results will be disseminated through user-friendly software tools and a summary statistics database. This work will help researchers better utilize various genetic markers, rich omics data, and environmental variables for a more comprehensive and unbiased understanding of how molecular changes contribute to disease causalities, ultimately enhancing public health through better interventions and treatments. The candidate will receive training from a mentoring team of globally recognized experts in the fields of statistical genetics, machine learning, genomics, epidemiology, and subject-matter expertise in cardiovascular diseases, cancer, and mental health; supported by a vibrant intellectual environment at Johns Hopkins University with seminars, collaborations, career development resources, and advanced coursework. This award will allow the candidate to gain critical skills in research, mentoring, communication, and leadership that will ensure success in her long-term goal of establishing an independent research program, focused on pioneeri...