As it is based on gene-level statistics obtained in a meta-analytic sample, our study overcomes some limitations of previous exploratory pathway studies 10, 53, 58 such as e.g., the literature selection bias, the bias arising from the use of incongruent analysis protocols or the heterogeneity problem 56, 59, 60 arising from incongruent definition, assessment, and/or distributional properties of the phenotype in question across the selected studies. More importantly, we (and, previously, Vink et al. in 39) use a hypothesis-free/unbiased genome-wide approach in deriving the list of input genes to be included in the pathway enrichment analyses. In so doing, our study surmounts the ‘circularity’ bias of previous pathway studies10, 58 built on lists of input genes mostly derived from published confirmatory/candidate gene studies. Furthermore, by being based on HYST 14 – which employs the scaled chi-square test to combine the LD-block-based p-values calculated with the GATES procedure – the study presented herein overcomes the limitations of other pathway-based tests such as e.g., the overrepresentation tests which do not take into account the correlation structure among the genes within pathways and among the SNPs within genes, or require arbitrary p-value thresholds to select the list of input genes.
