Fig. 5

Result comparisons of regrouped data SmitsData 1 (A1–A4) and SmitsData 2 (B1–B4) with different levels of heterogeneity. A1 and B1 plot the U-Curve of number of significant hits. In A2 and B2, we rank the significance of the 786 features, and then plot their \(-\,\log _{10} p\)-values with the same order across permutation scenarios. A3 and B3 plot the U-Curve of proportion of significant hits. A4 and B4 plots the fragility index of the top 100 features with a decreasing order. To save space, the legend listing the names of the 100 features are omitted