Archipelago method for variant set association test statistics

Variant set association tests (VSAT), especially those incorporating rare variants via variant collapse, are invaluable in genetic studies. However, unlike Manhattan plots for single-variant tests, VSAT statistics lack intrinsic genomic coordinates, hindering visual interpretation. To overcome this, we developed the Archipelago method, which assigns a meaningful genomic coordinate to VSAT P values so that both set-level and individual variant associations can be visualised together. This results in an intuitive and information rich illustration akin to an archipelago of clustered islands, enhancing the understanding of both collective and individual impacts of variants. In a validation study using 504 East Asian subjects from the 1000 Genomes Project phase 3 (250 cases, 254 controls), we first conducted a single-variant GWAS using logistic regression with Firth correction. We then performed an equivalent case/control analysis by collapsing variants into protein pathways (defined via STRINGdb and PreoMCLustR) and testing associations with SKAT-O. The Archipelago plot is applicable in any genetic association study that uses variant collapse to evaluate both individual variants and variant sets, and its customizability facilitates clear communication of complex genetic data. By integrating at least two dimensions of genetic data into a single visualization, VSAT results can be easily read and aid in identification of potential causal variants in variant sets such as protein pathways.