The enrichment score is expected to be high if most SNPs within a pathway are at the top of the list. We examine the statistical significance of a pathway by a permutation procedure. In each permutation, we permute (i.e. exchange) the phenotype labels, re-compute the P-values for SNPs and the corresponding enrichment score (denoted as perm_ES). Due to the size of large-scale genetic data, computational complexity would become extremely high when the number of permutations is very large. We used 1,000 permutation-cycles to generate the permutated datasets. The nominal P-value is obtained by comparing the enrichment score for the observed phenotypes with scores computed from permutated phenotypes.
