Table 4. Average Metrics for Each of Five Validation Folds from Cross-Validation Using the Full Set of 6955 Molecules Annotated with Only One of the 12 MeSH Classes.

problem group	metric	IMG + CNN	MFP + RF
3-class	accuracy	0.884 ± 0.0108	0.882 ± 0.0142
	balanced accuracy	0.879 ± 0.0143	0.870 ± 0.0162
	MCC	0.823 ± 0.0168	0.822 ± 0.0217
	ROC AUC	0.970 ± 0.0063	0.978 ± 0.00382
	ave. precision score	0.950 ± 0.0108	0.978 ± 0.00382
5-class	accuracy	0.863 ± 0.0104	0.871 ± 0.00700
	balanced accuracy	0.828 ± 0.0167	0.822 ± 0.0183
	MCC	0.811 ± 0.0140	0.821 ± 0.00969
	ROC AUC	0.972 ± 0.0046	0.981 ± 0.00284
	ave. precision score	0.933 ± 0.0093	0.950 ± 0.00582
12-class	accuracy	0.834 ± 0.0084	0.838 ± 0.00677
	balanced accuracy	0.735 ± 0.0258	0.719 ± 0.0248
	MCC	0.793 ± 0.0105	0.797 ± 0.00831
	ROC AUCa	0.969 ± 0.0026	0.977 ± 0.00227
	ave. precision scorea	0.900 ± 0.0073	0.918 ± 0.00392

^aReceiver operator characteristic area under the curve (ROC AUC) and average precision score were computed as the weighted average of scores across classes.