Fig. 2

Comparison between local/global graph learning and their combination. a shows the AUC and AUPR achieved by iPiDA-LGE with different fusion coefficients on \({\mathbb{S}}_{\text{validation}}\). b and c show the ROC and PR curves obtained by different predictors on \({\mathbb{S}}_{\text{validation}}\), respectively. d–f show the binary distribution of true labels and association scores predicted by iPiDA-L, iPiDA-G, and iPiDA-LGE on \({\mathbb{S}}_{\text{validation}}\)