摘要
Over the past two decades, genome-wide association studies (GWAS) have been successful in identifying robust associations between single nucleotide polymorphisms (SNPs) and the phenotypes of interest. The availability of large-scale GWAS data for complex traits/diseases have laid the foundation for drug development [1, 2] and precision medicine [3, 4]. Despite the great achievements, Non-European populations are severely under-represented in GWASs. According to the GWAS Diversity Monitor (https://gwasdiversitymonitor.com), about 88.9% of GWAS participants have been of European ancestry (EUR) to date. The lack of ancestral diversity in GWAS leads to bias in the understanding of the biology of complex traits [5, 6] and translating the GWAS findings into clinical applications in the under-represented populations (e.g., East Asian, African, etc.). In this thesis, we proposed two statistical methods to address challenges in these two directions. Given the fact that intensifying data-collection efforts are required to fill the disparities gap of genetic studies between non-Europeans and European, recent efforts have been devoted to improving the statistical power of GWASs in non-Europeans by applying trans-ancestry association mapping (TRAM) to integrate GWAS datasets across multiple populations. However, The challenges of TRAM arise from two major aspects. First, the genetic architectures of a phenotype are heterogeneous across ancestries. Some trait-associated SNPs have vastly different allele frequencies between European and non-European ancestries. SNP effect sizes and linkage disequilibrium (LD) patterns can also vary across ancestries. Second, the publicly released GWAS summary statistics still suffer from confounding biases. Although principal component analysis (PCA) and linear mixed models (LMMs) are commonly used for association mapping in GWASs, the population stratification in biobank-scale data, such as socioeconomic status or geographic structure, may not be fully accounted for in these standard approaches. Without correcting confounding biases hidden in GWAS summary statistics, TRAM will produce many false positive findings. As increasing sample sizes from GWASs, polygenic risk scores (PRS) constructed by a collective contribution of SNPs across the genome from GWAS has shown great potential in personal and clinical utility for a number of heritable diseases, including aid diagnosis by stratifying patients into different risk groups, early and cost-effective interventions, and improved therapeutic strategies. However, because of the aforementioned heterogeneous genetic architectures (e.g., allele frequencies, effect sizes, and LD) across different populations, the PRS constructed using European samples becomes less accurate when it is applied to individuals from non-European populations. It is an urgent task to improve the accuracy of PRSs in the under-represented populations. In this thesis, we propose a statistical method, LOG-TRAM, to leverage the local genetic architecture for TRAM. By using biobank-scale datasets, we showed that LOG-TRAM can greatly improve the statistical power of identifying risk variants in under-represented populations while producing well-calibrated p-values. We applied LOG-TRAM to the GWAS summary statistics of various complex traits/diseases from Biobank Japan, UK Biobank and African populations. We obtained substantial gains in power and achieved effective correction of confounding biases in TRAM. We also showed that LOG-TRAM can be successfully applied to identify ancestry-specific loci and the LOG-TRAM output can be further used for construction of more accurate polygenic risk scores in under-represented populations. Besides, to improve the accuracy of PRSs in trans-ancestry analysis, we also propose a cross-population and cross-phenotype (XPXP) method for construction of PRSs in under-represented populations. XPXP can construct accurate PRSs by leveraging biobank-scale datasets in European populations and multiple GWASs of genetically correlated phenotypes. XPXP also allows to incorporate population-specific and phenotype-specific effects, and thus further improves the accuracy of PRS. Through comprehensive simulation studies and real data analysis, we demonstrated that our XPXP outperformed existing PRS approaches. We showed that the height PRSs constructed by XPXP achieved 9% and 18% improvement over the runner-up method in terms of predicted R2 in East Asian and African populations, respectively. We also showed that XPXP substantially improved the stratification ability in identifying individuals at high genetic risk of Type 2 Diabetes.