Comparison of true positive and false positive rates to detect the 10% of main effect genes for different fold changes in 100 replicate simulated data sets. In each plot, the fold change increases from left to right. True positive rates are on top panels and false positive rates are on bottom panels. Sample sizes are M = 20 (left panels) and M = 40 (right panels). We use Gaussian mixture modeling (GMM) to determine the statistical thresholds for Relief-F, Random Forest, reGAIN plus CINC and dcGAIN plus CINC.