Prediction of protein long-range contacts using an ensemble of genetic algorithm classifiers with sequence profile centers

Table 2 Performance comparison on CATH classes

CATH Class	Sequence Length	Classification Accuracy (%)				ratio*	Protein Number
		2L	L	L/2	L/5
Alpha	<100	6.1	10.69	18.6	30.67	2.53	14
	100-200	4.57	7.41	10.44	23.43	1.53	30
	>200	5.34	7.32	11.06	16.74	0.96	27
	Average	5.16	8.02	12.28	22.31	1.51	71
Beta	<100	8.9	10.66	17.18	35.44	6.03	14
	100-200	5.72	8.33	13.5	30.81	2.68	56
	>200	5.03	7.62	13.2	28.23	1.95	35
	Average	5.92	8.41	13.89	30.57	2.88	105
Alpha Beta	<100	7.39	7.81	13.98	25.98	5.56	30
	100-200	4.77	6.93	10.63	24.11	2.23	99
	>200	3.8	5.54	8.75	15.69	1.39	112
	Average	4.65	6.39	10.17	20.43	2.25	241
Few SS**	<100	9.03	13.89	14.71	31.71	2.02	1
	100-200	5.43	9.15	14	14.29	1.2	3
	>200	4.92	7.92	10.67	13.79	0.68	2
	Average	5.86	9.53	13.01	17.03	1.16	6
Multi-domain chains	<100	4.71	7.69	10.35	12.5	2.3	3
	100-200	3.42	4.88	8.25	9.34	1.58	28
	>200	3.21	4.77	7.18	7.7	1.14	26
	Average	3.39	4.98	7.88	8.76	1.42	57
*All****	<100	7.34	9.2	15.58	28.61	4.77	62
	100-200	4.82	7.12	11.09	23.7	2.15	216
	>200	4.16	6.06	9.65	16.96	1.39	202
	Average	4.87	6.94	11.06	21.49	2.17	480

*The ratio of the number of residue pairs in long-range contact to that of total long-range residue pairs.
**Protein chains containing few secondary structures.
***All protein chains in our dataset.