Table 4.

Five-fold cross-validation results of the models trained with various features for classifying between 96 carbonylated and 488 non-carbonylated threonine residues

Classifier	Training features	Sensitivity	Specificity	Accuracy	MCC
SVM	AA	0.625	0.615	0.616	0.180
	AAC	0.667	0.656	0.658	0.244
	AAPC	0.646	0.660	0.658	0.232
	PWM	0.688	0.672	0.675	0.274
	PSSM	0.656	0.656	0.656	0.236
	ASA	0.573	0.590	0.587	0.122
	AAindex	0.667	0.654	0.656	0.242
J48 DT	AA	0.604	0.594	0.596	0.148
	AAC	0.635	0.635	0.635	0.204
	AAPC	0.635	0.641	0.640	0.209
	PWM	0.625	0.637	0.635	0.198
	PSSM	0.604	0.598	0.599	0.151
	ASA	0.573	0.590	0.587	0.122
	AAindex	0.646	0.641	0.642	0.217
RF	AA	0.625	0.617	0.618	0.181
	AAC	0.656	0.652	0.652	0.233
	AAPC	0.646	0.652	0.651	0.225
	PWM	0.677	0.668	0.670	0.262
	PSSM	0.656	0.656	0.656	0.236
	ASA	0.583	0.594	0.592	0.133
	AAindex	0.656	0.676	0.673	0.254

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