Table 2.

Five-fold cross-validation results of the models trained with various features for classifying between 206 carbonylated and 1166 non-carbonylated lysine residues

Classifier	Training features	Sensitivity	Specificity	Accuracy	MCC
SVM	AA	0.680	0.643	0.649	0.235
	AAC	0.728	0.686	0.692	0.305
	AAPC	0.699	0.696	0.697	0.294
	PWM	0.748	0.715	0.720	0.346
	PSSM	0.704	0.686	0.689	0.288
	ASA	0.592	0.571	0.574	0.117
	AAindex	0.709	0.720	0.719	0.323
J48 DT	AA	0.534	0.557	0.554	0.066
	AAC	0.655	0.678	0.674	0.246
	AAPC	0.670	0.683	0.681	0.261
	PWM	0.689	0.674	0.676	0.267
	PSSM	0.621	0.660	0.655	0.207
	ASA	0.515	0.563	0.555	0.055
	AAindex	0.660	0.682	0.679	0.253
RF	AA	0.660	0.635	0.638	0.214
	AAC	0.704	0.686	0.689	0.288
	AAPC	0.709	0.703	0.704	0.307
	PWM	0.718	0.707	0.708	0.317
	PSSM	0.699	0.686	0.688	0.285
	ASA	0.583	0.583	0.583	0.119
	AAindex	0.709	0.717	0.716	0.319

The numbers marked with italicized font are the highest values in four measurements