Table 2. Performance comparisons between EPMDA and seven existing computational models (RLSMDA, HMDP, WBSMDA, MCMDA, HGIMDA RWRMDA and RBMMMDA) for predicting microRNA-disease association in terms of AUCs based on leave-one-out and 5-fold cross validations. All the eight models adopt the disease semantic similarity based on disease MeSH annotations.

METHOD	LOOCV	5-fold cross validation
RLSMDA[38]	0.8426	0.6953
HDMP[42]	0.8366	0.7702
WBSMDA[37]	0.8030	0.8031
MCMDA[34]	0.8749	0.8767
HGIMDA[41]	0.8781	0.8077
RWRMDA[39]	0.8617	0.7891
RBMMMDA[40]	0.8606	N/A
EPMDA (The proposed method)	0.8945	0.8917+/−0.0004