Fig. 2.

Performance comparisons between NCPHMDA and three state-of-art prediction models (KATZHMDA, Regularized Least Squares and Random Walk with Reastart) in terms of ROC curve and AUC. As a result, NCPHMDA achieved AUCs of 0.9039 and 0.7953 based on global and local LOOCV, significantly outperforming previous classification models