Fig. 4

Leukemia gene expression results. Scatter plots showing the association between phenotype gene set enrichment p-values and PC gene set enrichment p-values for the Armstrong et al. [40] leukemia gene expression data, AML/ALL phenotype, MSigDB C2 v4.0 gene sets and first three PCs. Both phenotype and PC gene set enrichment p-values were computed as outlined in Section “Evaluation using MSigDB C2 v4.0 gene sets and Armstrong et al. leukemia gene expression data”. Shown in each plot is the Spearman correlation coefficient between phenotype and PC gene set enrichment p-values and the positive predictive value of PC gene set enrichment for identifying gene sets that are significantly enriched relative to the phenotype at an α=0.1 (shown by dotted lines). True positives are in the upper right quadrant, false positives are in the upper left quadrant. Plots a-c show the association between phenotype and PC gene set enrichment p-values for PCs 1 through 3 with the PC enrichment p-values computed using a two-sided t-test on the standardized mean difference gene set statistic. For plots d-f, the PC gene set enrichment p-values were computed using a correlation-adjusted two-sided t-test and, for plots g-i, the PC gene set enrichment p-values were computed using the permutation distribution of the gene set statistic