Protein local 3D structure prediction by Super Granule Support Vector Machines (Super GSVM)

BMC Bioinformatics

Table 1 The improvement of the number of high quality sequence clusters in each information granule

	G0	G1	G2	G3	G4	G5	G6	G7	G8	G9	Total
Total number of clusters	151	76	95	72	70	133	143	5	48	6	799
Original FGK-250
60%~70%	36	24	24	28	32	31	35	2	15	4	231
70%~80%	21	3	12	4	4	24	20	0	0	0	88
>80%	7	0	7	0	0	4	6	0	0	0	24
Super GSVM
60%~70%	44	30	31	30	42	39	40	3	24	4	287
70%~80%	27	17	17	16	16	27	30	1	4	1	156
>80%	26	2	19	2	1	26	23	0	0	1	100

In this work, we use Super GSVM to generate and extract sequence clusters. We divided the whole training dataset into 10 information granules, the second row shows the number of clusters in each information granule (detail of this information can be found in [11]). The table also show the number of clusters belong to excellent (>80% 2^nd structural similarity), good (70%~80% 2^nd structural similarity), and fair (60%~70% 2^nd structural similarity) clusters before and after the Super GSVM extraction.

ISSN: 1471-2105