BacHbpred: Support Vector Machine Methods for the Prediction of Bacterial Hemoglobin-Like Proteins

Table 1

Performance of various SVM modules of HbL proteins predictions with non-HbL and HbL classification (single domain, two domains (flavoHbs) and truncated Hbs (trHb)) developed using various methods: amino acids (AC), dipeptides (DC), PSSM, and MM profiles.

	Methods	ACC	SN	SP	MCC	Parameter
HbL versus non-HbL	AC	86.14	96.18	76.11	0.82	25	400
	DC	83.02	94.78	71.27	0.78	1	375
	PSSM	90.20	97.76	82.64	0.89	1	300
	MM	86.28	96.08	76.49	0.83	25	450
	Hybrid	85.21	95.80	74.62	0.81	0.1	375
sHb	AC	94.96	100	94.56	0.97	15	9
	DC	83.23	100	82.15	0.91	0.2	250
	PSSM	95.05	100	94.66	0.97	5	7
	MM	94.87	100	94.46	0.97	1	150
	Hybrid	91.51	100	90.83	0.95	0.1	350
FlavoHb	AC	96.46	100	89.67	0.95	10	300
	DC	87.50	100	63.58	0.80	1	350
	PSSM	95.05	100	85.59	0.93	1	350
	MM	96.46	100	89.67	0.95	10	300
	Hybrid	90.29	100	71.73	0.84	1	150
trHb	AC	85.26	98.89	80.62	0.89	5	350
	DC	78.17	98.89	71.13	0.83	1	275
	PSSM	87.97	100	83.88	0.92	1	400
	MM	85.07	99.26	80.25	0.89	4	500
	Hybrid	80.03	100	73.25	0.85	1	150