JasMAP: A Joint Ancestry and SNP Association Method for a Multi-way Admixed Population

The large volume of research findings submitted to the GWAS catalog in the last decade is a clear indication of the exponential progress of these studies and association approaches. This success has, however, been dimmed by recurring concerns about disparity and the lack of population diversity. As a result, researchers are now responding, and GWAS extension to diverse populations is under way. Initial GWAS methods were calibrated using European populations with long-range regions of linkage disequilibrium (LD) and haplotypes. This implies that, as GWAS extends to diverse populations, the development of inclusive methods targeted at these populations is imperative. Particularly in multi-way admixed populations, methods that include both genotypes and ancestry associations have been shown to improve power while controlling for the additional LD structure introduced by admixture processes. However, these methods continue to be tailored to only 2-way admixed populations. Though this is a justifiable start, the breeding structures of today suggest that the world population is more likely to increase in the number of multi-admixed individuals, and tools targeted at 2-way admixed individuals will continue to exclude a larger part of diverse populations. In this study, we propose a joint ancestry and SNP association method, JasMAP, that is tailored to multi-way admixed populations. We explore the LMM approach that has become standard in GWAS of structured populations in a Bayesian context, model local ancestry variation as prior knowledge, and update the genotype association to obtain a joint posterior probability of association (PPA). The newly developed method has been assessed using various simulated datasets from our multi-scenario simulation framework, FractalSIM (Mugo et al., 2017), and we output not only the joint statistics but also the genotype-only and the ancestry-only association statistics for the user. JasMAP has also been applied to perform a GWAS analysis of a 5-way admixed South African Coloured (SAC) population with a tuberculosis (TB) phenotype. We obtained 1 significant risk SNP using the ancestry-only association but no SNPs were found to be significant using the standard genotype-only association. 13 risk SNPs, however, were detected as significant with a PPA > 0.5 using the joint association approach. 12 of these SNPs had a marginal significance threshold in genotype-only and ancestry-only association. By functional annotation and gene mapping, the 13 SNPs were found near 8 genes, 5 of which were either found in pathways, have functionality, or were linked to social behaviour associated with an increased risk of TB. Specifically, one of the significant SNPs, rs17050321 on chromosome 4, was found close to the SLC7A11 gene that has previously been linked to TB in a GWAS study of a Chinese population.