Combination of scoring schemes for protein docking

BMC Bioinformatics

Table 2 Combination of scoring schemes

	unopt	amino acid × atom specific	SVM	AA × ATM & SVM	unopt	amino acid × atom specific	SVM	AA × ATM & SVM	unopt	amino acid × atom specific	SVM	AA × ATM & SVM
Validation (Literature-data, 12°)
	Enzyme-Inhibitor (21)				Antibody-Antigen (4)				Others (4)
No on 1	1	4	1	5	0	0	1	0	1	0	1	0
No = 10	2	7	7	13	0	0	1	0	1	0	1	0
No = 50	6	12	16	15	1	0	2	2	1	0	1	0
No = 100	8	13	18	17	1	0	2	2	1	0	1	0
Average	706	606	58	55	9887	442	299	115	3493	4188	1978	2701
Validation (Benchmark 2.0, 12°)
	Enzyme-Inhibitor (22)				Antibody-Antigen (19)				Others (29)
No on 1	0	2	4	3	0	1	0	1	0	0	0	1
No = 10	0	5	11	11	1	2	3	4	1	4	0	6
No = 50	1	12	18	19	1	3	7	10	2	7	5	7
No = 100	7	15	18	19	1	7	9	12	3	9	7	12
Average	897	416	113	104	5393	1245	350	302	4846	1219	1117	712

Number of complexes for which a near-native structure can be predicted on the first and within the top10, top50 and top100 ranks. Furthermore the average rank of the first near-native structure is given. These quality measurements are shown for an non-optimised ranking, based on the geometric correlation (unopt), a ranking based on the combination of amino-acid specific and atom specific weighting factors (amino acid × atom specific), a ranking based on SVM scoring (SVM), and for the combination of amino acid, atom specific and SVM-based ranking AA × ATM & SVM. The combination of all three scoring schemes for enzyme-inhibitor complexes and for antibody-antigen complexes is done by ranking all structures by the SVM score and than ranking all structures with the same score again by the weighted geometric correlation. For the 'other' complexes the combination is done by multiplying all scores.

ISSN: 1471-2105