Multi-ancestry colocalization approaches

Genome-wide association studies (GWAS) have identified thousands of variants associated with complex traits, but many are non-causal. Statistical fine-mapping methods aim to pinpoint the most likely causal variants among the many associated ones. While most fine-mapping methods were originally limited to single ancestry analysis, multi-ancestry fine-mapping methods are now available, leveraging differences in linkage disequilibrium (LD) and minor allele frequencies (MAFs) across ancestries to improve fine-mapping resolution. However, the biological relevance of the putative causal variants identified through fine-mapping often remains unclear. Colocalization methods improve interpretability by integrating GWAS data with other functional genomics datasets to assess whether two traits share the same causal variants. Despite the growing availability of multi-ancestry data, there are currently no established methods for multi-ancestry colocalization. In this study, we propose multi-ancestry colocalization approaches through the integration of multi-ancestry fine-mapping methods, SuSiEx and MsCAVIAR, with single ancestry colocalization methods, coloc and eCAVIAR. We introduce coloc_SuSiEx, eCAVIAR_SuSiEx, eMsCAVIAR and coloc_MsCAVIAR. The performance of the proposed approaches is evaluated and compared through simulation studies. In loci with a single causal variant, credible set sizes across the four approaches were comparable, as was the prioritization of the true causal variant. MsCAVIAR-based approaches were more computationally expensive compared to SuSiEx-based approaches, which is an important consideration for the analysis of regions with multiple causal variants. Compared to the coloc-based approaches, the eCAVIAR-based approaches tended to report lower loci level colocalization posterior probabilities. For the analysis of loci with multiple causal variants, coloc_SuSiEx is the preferred approach. We apply the proposed approaches to perform a colocalization analysis of multi-ancestry T2D GWAS data from the DIAMANTE Consortium and European pQTL data from the INTERVAL study. This work addresses the increasing need for multi-ancestry approaches to colocalization analysis as more multi-ancestry data become available.Author summary: Through genome-wide association studies (GWAS), thousands of genetic variants associated with complex traits have been identified. Pinpointing the causal variants among the numerous associated ones is a challenge that statistical fine-mapping methods aim to solve. Colocalization is another common downstream analysis and it assesses whether two traits share the same causal variant. As data from diverse ancestries become more widely available, there is a growing need for multi-ancestry methods. However, there are currently no established methods in the literature for multi-ancestry colocalization analysis. To bridge this gap, we introduce four multi-ancestry colocalization approaches: coloc_SuSiEx, eCAVIAR_SuSiEx, eMsCAVIAR and coloc_MsCAVIAR. These approaches are developed through the cross-integration of the multi-ancestry fine-mapping methods, SuSiEx and MsCAVIAR, with single ancestry colocalization methods, coloc and eCAVIAR. The performance of the proposed approaches is investigated through extensive simulation studies, and their applicability is demonstrated through a colocalization analysis of a multi-ancestry T2D GWAS dataset with a European pQTL dataset.