Next, we evaluated the type I error and power of GPA for hypothesis testing on the significance of annotation enrichment for risk SNPs. Gene Set Enrichment Analysis (GSEA) [34] is a popular method to accomplish a similar task. Although GSEA typically is used for gene expression data analysis, its input can be a list of p-values obtained from any source. Therefore we implemented the GSEA method to test the enrichment of the -values of a set of SNPs being annotated and compared it with GPA. We followed the previous simulation scheme and simulated one GWAS data set with , varying from 2000 to 10000, and varying from 500 to 2000. Here was fixed at 0.1 and was varied from 0.1 to 0.5. We set the statistical significance level at 0.05. Type I error rate was evaluated at and power was evaluated at . The results for are shown in Figure 4. In general, GPA provided much higher power than GSEA while both methods appropriately controlled the type I error rate.
