Most of the current methods for pathway analysis of GWAS data are gene-based. Wang et al. [13] used the SNP with the strongest association to represent a gene. Choosing the smallest P-value for each gene might not be optimal in situations when the joint action of multiple SNPs within a gene explains more variance than the most significant SNP. For example, if a gene contains multiple causal variants, it might not be identified by the smallest P-value method, which reduces the power of the subsequent pathway enrichment analysis. Moreover, this approach is likely to favour genes of large sizes, as genes with a larger number of SNPs have a higher chance of having significant SNPs, by chance alone. Consequently, the effects of genes with smaller numbers of SNPs would be underestimated. Holmans et al. [10] proposed ALIGATOR (Association LIst Go AnnoTatOR) method to study the significance of pathways. Although this method corrects variable gene sizes by simulations, it requires a pre-determined P-value cutoff for selecting significant SNPs and the evaluation of pathways is gene-based, not SNP-based. Yu et al. [18]
