Improving Methods and Practices for Trans-Ethnic Genetic Studies

摘要

ABSTRACT Trans-ethnic genetic analysis can facilitate the discovery of trait- or disease-associated loci, characterize shared and differential genetic architectures across populations, improve the delineation of causal variants, and is critical for equal delivery of genomic knowledge and precision healthcare globally. However, current trans-ethnic genetic research is impeded by (i) limited genomic resources for non-European populations; and (ii) limited statistical methods that can appropriately model and integrate data from diverse populations. This project will address these challenges by (1) aggregating and harmonizing genetic data, physical measures, laboratory tests and disease information from global biobanks and multiple health care systems in the United States, with >680K samples of non-European ancestry and a total sample size >1.4M by 2022; and (2) developing statistical methods and best practices to integrate multi-ethnic data for improved cross-population characterization of genetic architectures, meta-analysis, statistical fine-mapping and polygenic prediction. Specifically, in Aim 1, we will systematically characterize the comparative genetic architectures of physical measures, biomarkers and disease phenotypes at variant, locus and genome-wide levels within and across continental populations, and discover novel genetic loci through trans-ethnic meta-analysis. In Aim 2, we will develop novel statistical methods and establish best practices for trans-ethnic fine-mapping, delineate putative causal genetic variants for a range of complex traits and diseases, and explore the biological mechanisms of fine-mapped variants. In Aim 3, we will develop novel haplotype-based methods for trans-ethnic polygenic prediction, comprehensively assess the factors that might affect the transferability of polygenic risk scores (PRS) and benchmark the clinical utility of biomarker PRS in disease risk prediction across diverse populations. We are committed to resource sharing and will publicly release genome-wide association summary statistics, reference panels, fine-mapping results, and polygenic prediction pipelines produced in this project. All statistical methods and bioinformatic tools developed in this project will be disseminated as publicly available software packages.