Visualisation of variable binding pockets on protein surfaces by probabilistic analysis of related structure sets

Table 1 Ensemble analysis of the binding sites of protein-protein interaction inhibitors

Protein	Binding site size (atoms)	Method	Percentage of site identified as pocket-lining
			*apo* structure	Provar Score > 0.75	Provar Score > 0.25
		PASS	38	21	64
Bcl-X_L	56	LIGSITE	73	69	91
		fpocket	52	36	89
		PASS	46	15	50
MDM2	48	LIGSITE	52	38	77
		fpocket	38	6	45
		PASS	29	26	69
XiapBir3	42	LIGSITE	67	52	98
		fpocket	31	24	83
		PASS	25	16	54
XiapBir3	63	LIGSITE	57	44	87
		fpocket	30	19	74
		PASS	70	35	75
ZipA	20	LIGSITE	35	5	60
		fpocket	0	15	15
		PASS	46	23	54
HPVE2	97	LIGSITE	44	26	56
		fpocket	33	2	53
		PASS	38	7	58
IL2	60	LIGSITE	22	22	62
		fpocket	0	2	2
		PASS	53	53	68
HIV-1 Integrase	19	LIGSITE	37	16	47
		fpocket	68	5	58
		PASS	6	0	29
TNFa	34	LIGSITE	3	0	44
		fpocket	0	0	3
		PASS	0	0	69
TNFR1a	16	LIGSITE	50	38	100
		fpocket	0	0	69
		PASS	62	30	57
MDM4	81	LIGSITE	44	31	68
		fpocket	38	0	42
		PASS	38.5	20.5	59
Mean Values	49	LIGSITE	49	31	72
		fpocket	37.5	8.5	48

Provar analysis of a tCONCOORD generated ensemble of 250 structures of the apo form of each of the 11 proteins in the 2P2I database identifies a greater proportion of the known binding site atoms as pocket-lining than analysis of the crystallographic structure alone. The proportion of inhibitor binding-site atoms found to be persistently or variably lining pockets (defined as pocket-lining in at least 25% of conformers) is significantly greater than for the protein as a whole in 21/33cases. (An italic number in the final column indicates the minority of cases in which the proportion is not significant at the p < 0.05 level). Further details of the structures are given in Table 4.