Table 1. Performance of various prediction models on training dataset.

Features	MCC		Accuracy		Sensitivity		Specificity
	SVM	RF	SVM	RF	SVM	RF	SVM	RF
AAC	0.664	0.689	0.858	0.868	0.695	0.706	0.935	0.945
ATC	0.519	0.587	0.802	0.826	0.503	0.658	0.942	0.905
PCP	0.420	0.553	0.759	0.814	0.524	0.599	0.869	0.915
DPC	0.653	0.644	0.853	0.850	0.706	0.599	0.922	0.967
AAC+ATC+PCP+DPC	0.697	0.698	0.872	0.872	0.706	0.722	0.95	0.942
AAC+PCP+DCP	0.693	0.661	0.870	0.856	0.706	0.620	0.947	0.967
AAC+PCP+ATC	0.685	0.698	0.867	0.872	0.695	0.727	0.947	0.940
AAC+PCP	0.681	0.681	0.865	0.865	0.695	0.695	0.945	0.945
AAC+ATC	0.664	0.673	0.858	0.862	0.695	0.642	0.935	0.965
AAC+DCP	0.673	0.657	0.862	0.855	0.701	0.61	0.937	0.970
PCP+ATC+DCP	0.661	0.669	0.856	0.86	0.711	0.631	0.925	0.967
PCP+ATC	0.595	0.664	0.831	0.858	0.615	0.685	0.932	0.940
PCP+DCP	0.661	0.661	0.856	0.856	0.701	0.620	0.93	0.967
ATC+DCP	0.657	0.661	0.855	0.856	0.701	0.620	0.927	0.967

The first column represents the features. The second, the third, the fourth and the fifth respectively represent the MCC, accuracy, specificity and sensitivity. Columns 2-5 subdivided into two parts namely SVM- (normal font) and RF-based (underlined) performances. AAC: amino acid composition; ATC: atomic composition; PCP: physiochemical properties; DPC: dipeptide composition. Features that gave the highest MCC is shown in bold.