Fig. 3

Overlaps (expressed as Jaccard indices) between true positive predictions of the eight considered (sub)networks, each under its best predictive model, when using the primary GenAge-based definition of aging- and non-aging-related gene labels, and when using HPRD. On the x- and y-axes, the numbers in the parentheses after the name of each subnetwork are the number of true positive predictions (i.e., known aging-related genes) and the number of total predictions. Within each table/matrix cell, the number on the first line is the Jaccard index of the overlap between the corresponding pair of subnetworks; the number on the second line is the raw overlap size, i.e., the actual count of true positives that are in the overlap between the two subnetworks; the number on the third line is the adjusted p-value of the overlap with respect to the hypergeometric test. The darker the color of a given cell, the larger the Jaccard index value, i.e., the higher the overlap. Note that we have two groups of results for wNetWalk-Dynamic, i.e., wNetWalk-Dynamic Existing and wNetWalk-Dynamic Proposed, which we already explained in the caption of Fig. 2. For results when using BioGRID instead of HPRD, see Additional file 1: Fig. S5