Table 1. Performance of SVM classifiers for various combinations of protein sequence features, kernels, parameters and validation methods.

Model	Feature	Dm	Validation	ACC (%)	SN (%)	SP (%)	MCC	F1	Parameters
									C	γ
Module1	AAC	20	a	83.57	83.82	83.33	0.671	0.712	1.5	84
Module2	DPC	400	a	85.13	85.29	84.72	0.699	0.734	30	12
Module3	SSC	60	a	80.71	76.47	84.72	0.614	0.658	7	10
Module4	PSSM	400	a	92.14	92.64	91.66	0.842	0.851	32.5	1
			b1	97.85	97.05	98.61	0.957	0.889	47.5	0.5
			b2	100	100	100	1	1	0.5	100
			c1	i) 88.57	88.23	88.88	0.771	0.787	4	9.6
			c2	ii)94.28	91.17	97.22	0.886	0.872	19	0.5
Hybrid-1	AAC+DPC	420	a	81.42	82.35	80.55	0.628	0.682	880	0.1
Hybrid-2	AAC+SSC	80	a	85.71	89.70	81.94	0.717	0.753	10.8	10
Hybrid-3	DPC+SSC	460	a	81.42	79.41	83.33	0.628	0.675	10.8	10
Hybrid-4	PSSM+AAC	420	a	91.42	92.64	90.27	0.828	0.84	230	0.1
Hybrid-5	PSSM+DPC	800	a	91.42	92.64	90.27	0.828	0.84	30	1
Hybrid-6	PSSM+SSC	460	a	92.14	92.64	91.66	0.842	0.851	19.2	1.5
			b1	98.57	97.05	100	0.971	0.904	19.2	1.5
			b2	100	100	100	1	1	0.5	50
			c1	i) 88.57	88.23	88.88	0.771	0.787	0.5	30
			c2	ii)88.28	82.35	86.11	0.685	0.753	3	7
Hybrid-7	PSSM+DPC+SSC	860	a	90.0	91.17	88.88	0.80	0.815	10	2
Hybrid-8	PSSM+DPC+AAC	820	a	91.42	92.64	90.27	0.828	0.84	200	0.1
			b1	95.71	97.10	94.46	0.914	0.847	200	0.1
			b2	100	100	100	1	1	0.5	100
			c1	i) 85.71	73.52	97.23	0.731	0.862	0.5	19
			c2	ii)91.42	97.05	86.11	0.834	0.774	0.3	4

Dm: dimension, a = Jackknife test CV, b1 = self-consistency test (mode 1), b2 = self-consistency test (mode 2), c1 & c2 = holdout-test, SN: sensitivity, SP: specificity, MCC: Mathew's Correlation Coefficient, F1: F1 statistics, C: tradeoff value, γ: gamma factor (a parameter in RBF kernel).