Figure 3

Average results from the case i. Figure represents the AUC score for each evaluated method as the rank limit of the positive GO classes is increased. The set of positive classes used for AUC grows as the rank threshold becomes bigger. Methods represented are GSZ-score: blue line with circles, t-test: green line with cross, KS test: red line with box, modKS test: cyan line with diamond, iGA: magenta line with x. Lower part zooms into the smallest ranks. Here GSZ-score shows the best performance and t-test performs equally well with the top ranks, while other methods show weaker performance.