Table 2 The prediction performance of the Met-predictor based on 5-fold cross-validation on the training set.

Residue	Methods	Window Size	MCC	ACC	SEN	SPE	PRE	CPRE	AUC	PRAUC
K	Met-predictor(seq)	17	0.261	0.630	0.655	0.606	0.624	0.061	0.679	0.669
	MONO	23	0.206	0.674	0.927	0.215	0.682	0.682	0.674	0.789
	DI	17	0.384	0.692	0.674	0.709	0.699	0.421	0.752	0.746
	TRI	21	0.455	0.726	0.769	0.684	0.709	0.320	0.779	0.746
	Met-predictor(seq + str)	17	0.286	0.643	0.662	0.624	0.638	0.064	0.692	0.685
	MONO	23	0.207	0.674	0.925	0.219	0.683	0.683	0.676	0.792
	DI	15	0.390	0.695	0.686	0.703	0.698	0.420	0.756	0.724
	TRI	23	0.359	0.679	0.692	0.667	0.675	0.287	0.754	0.749
R	Met-predictor(seq)	17	0.371	0.685	0.654	0.716	0.697	0.089	0.749	0.759
	MONO	17	0.148	0.703	0.992	0.056	0.701	0.701	0.606	0.761
	DI	19	0.377	0.721	0.409	0.909	0.729	0.729	0.745	0.665
	Met-predictor(seq + str)	17	0.380	0.689	0.642	0.737	0.709	0.094	0.752	0.763
	MONO	17	0.184	0.709	0.981	0.101	0.709	0.709	0.636	0.774
	DI	21	0.302	0.692	0.326	0.912	0.690	0.690	0.712	0.607

Two versions of Met-predictor are included here: Met-predictor(seq), where only sequence-based features are used to build models, and Met-predictor(seq + str), which uses not only sequence-based features, but also the novel structure model-based features to build models. The lysine residue is represented by “K” while the arginine corresponds to “R”. The definitions of measures MCC, ACC, SEN, SPE, PRE, CPRE, AUC and PRAUC are shown in “Performance Evaluation” section and Eqs. (1) to (6).

Quick links

Search