Table 2.

Performance Comparison of Various Single Method-Based Models and a Selected Ensemble Model for Predicting m6A Sites on the Independent Dataset

Method	MCC	ACC	SN	SP	AUC
1. RF	0.715	0.858	0.842	0.873	0.923
2.ERT	0.729	0.864	0.855	0.873	0.934
3. SVM	0.742	0.871	0.873	0.869	0.936
4. XGB	0.721	0.860	0.891	0.828	0.939
{1, 2}	0.726	0.862	0.896	0.828	0.931
{2, 3}a	0.769	0.885	0.878	0.891	0.938
{3, 4}	0.757	0.878	0.905	0.851	0.940
{1, 3}	0.743	0.871	0.891	0.851	0.935
{1, 4}	0.731	0.864	0.905	0.824	0.936
{2, 4}	0.731	0.864	0.905	0.824	0.938
{1, 2, 3}	0.751	0.876	0.887	0.864	0.936
{2, 3, 4}	0.744	0.871	0.905	0.837	0.939
{1, 3, 4}	0.760	0.880	0.891	0.869	0.938
{1, 2, 4}	0.735	0.867	0.905	0.828	0.936
{1, 2, 3, 4}	0.731	0.864	0.905	0.824	0.936

The first column represents a single method-based model or an ensemble model, which was built based on combining different single models (see Table 1 legend for more information).

^aThe best performance obtained by the optimal model.