Fig. 4.

Top predicted entities co-occur with query terms in literature. We validated the top 20 predictions based on 2016 networks for each chosen query term using literature-based associations. We chose 10 diseases of leading cause of mortality as query terms to predict (A, D) associated genes and (B, E) FDA-approved drugs. (C, F) We also chose 10 genes associated with those 10 diseases as query terms and predicted associated FDA-approved drugs. Random control was 50 average experiments of 20 random terms for each query term. We computed (A–C) the precision at 20 and (D–F) chi-square test to determine how different the predictions from random controls. N.s., not significant (i.e. chi-square test P-value >0.05). The 1 mode, (A) DG, (B) DC, (C) GC network; 3 modes, DG+DC+GC network; 6 modes, DG+DC+GC+DD+GG+CC network