Table 1.

The performance of SVM models developed using different types of compositions. These models were trained and tested on DNAset, a dataset of DNA-binding and non-binding protein domains/chains.

Composition Type	Threshold	Sensitivity	Specificity	Accuracy	MCC
Amino Acids	0.3	78.11	80.80	79.80	0.58
Dipeptides	0.0	73.35	77.60	76.01	0.50
4-parts amino acids	0.0	77.45	77.60	77.53	0.54
PSSM	0.1	86.32	86.80	86.62	0.72