Receiver operative characteristic (ROC) curve performance by biomolecular interactions of pathophenotypic similarities.

<p>Physical interactions (dashed blue line), metabolic flux correlations (dashed green line), functional interactions (red continuous line) and an integrated interactome generated by the sum of all other interactomes (black continuous line). ROC curves were computed to assess the signal of pathophenotypic similarities for biological interactions. True positives (TP) were those interactions that where found in the intersection between PSGN and each biomolecular interactome (PIN, MGN and FSGN). The dataset of false positives (FP) was calculated from intersected gene pairs between PSGN and randomizations of each biomolecular interactome. We obtained severals different FP datasets to calculate the average area under the curve (AUC), it was 0.77 for PIN, 0.76 for MGN, 0.66 for FSGN and 0.68 for the integrated interactome. Only biochemical interactomes show significantly different AUCs to that of the integrated interactome (average p-values of 2.2E−6 and 4.1E−2 for PIN and MGN respectively).