A graph kernel method for DNA-binding site prediction

BMC Systems Biology

Table 2 Predictions by the top 1 patch.

Unbound PDB id ¹	Bound PDB id	Top 1 patch		P_Random(%) ⁴
		TP²
1iknA,C	1leiA,B	6	0	0.4
1g6nA,B	1zrfA,B	6	0	2.4
1zzkA	1zziA	6	0	2.9
1mml	1ztwA	4	2	4.2
1a2pC	1brnL	6	0	4.5
1ko9A	1m3qA	6	0	4.9
1qc9A	1cl8A,B	4	2	8.0
1lqc	1l1mA,B	6	0	8.7
1qzqA	1rfiB	2	4	9.5
1xx8A	1xyiA	6	0	10.7
1qqiA	1gxpA,B	4	2	19.3
1l3kA	1u1qA	3	3	25.6
2alcA	1f5eP	4	2	26.7

¹ For the proteins in italic, the interfaces were predicted using only PSSM. For the others, all seven features were used; ² TP: the number of the interface residues falling in the top 1 patch; ³ FP: the number of non-interface residues in the top 1 patch. ⁴ P_random: When a patch is randomly picked, the probability of it containing at least as many interface residues as the top 1 patch.

ISSN: 1752-0509