This simulation framework was also used to choose an optimal gene score enrichment cutoff, for our GSEA algorithm. Five cutoffs were tested: 99th, 95th, 90th, 75th, and 50th percentile of all gene p-values for two effect sizes: NCP = 2.5 and NCP = 10, assuming a total of 100 causal genes in the genome. A equivalent to the 95th percentile of for all genes g in the genome yielded the optimal power, when considering power plots for both effect sizes (Figure S5). The 75th percentile cutoff performed a bit better than the 95th percentile cutoff for very weak effects (NCP = 2.5; Figure S5B), especially when assuming a total of 500 causal genes (data not shown). Hence, the 75th percentile cutoff could be used for diseases or traits that are highly polygenic with many associations of weak effects.
