Fig. 4. Correlation between prediction scores and gene multifunctionality.

Boxplot of distance correlation between gene set prediction scores of genes (n = 19,635) and the gene set collection multifunctionality score (y-axis) for each gene set collection (x-axis). Multifunctionality scores are compared against ICA-TC-based (blue) or PCA-TC-based (red) predictions. The magnitude of correlation varies across gene set collections and between gene sets in a collection. The associations between the degree of multifunctionality and the prediction scores were lower for the ICA-TC-based method than the PCA-TC-based method (area under the curves (AUCs) calculated from two-sided Mann–Whitney U test range 0.03–0.47). Hinges of boxes represent second and third quartiles and whiskers extend by half that interquartile range. Center of box corresponds to median.