Table 3. Prediction performance of different algorithms on the experiment-validated miRNA: Gene interactions.

Algorithm	#predicted positive	#predicted negative	Sen.(%)	Spe.(%)	F1(%)	MCC(%)	Acc.(%)
cnnMirTarget	5179	162	66.27	57.65	78.99	9.21	65.97
miRTarget3	1000	263	12.80	93.59	22.64	3.53	15.60
metaMIR	5131	72	65.66	25.62	78.01	-3.37	64.27
TargetScan7	1455	272	18.62	96.80	31.36	7.33	21.33

The experiment-validated positive dataset contains 7815 interactions, while the negative dataset contains 281 pseudo-interactions. Column 2: number of predicted positive interactions in the positive dataset for each algorithm. Column 3: number of predicted negative interactions in the negative dataset for each algorithm. Column 4–8: Sensitivity (Sen.), Specificity (Spe.), F1-Score, MCC, Accuracy (Acc.).