Improving the prediction of protein binding sites by combining heterogeneous data and Voronoi diagrams

BMC Bioinformatics

Table 1 AUC values for different combinations of features and environment definitions

Features	Voronoi Diagrams	Sphere	Sliding Window	Single (no environment)
s	0.79	0.75	0.77	0.72
e	0.77	0.72	0.75	0.71
c	0.76	0.74	0.72	0.65
b	0.74	0.71	0.69	0.61
s+e	0.78	0.75	0.77	0.73
s+c	0.82	0.78	0.81	0.77
s+b	0.79	0.75	0.77	0.73
e+c	0.81	0.75	0.77	0.76
e+b	0.77	0.73	0.75	0.72
c+b	0.76	0.74	0.72	0.68
s+e+c	0.82	0.78	0.81	0.77
s+e+b	0.79	0.75	0.77	0.73
s+c+b	0.82	0.78	0.8	0.78
e+c+b	0.81	0.75	0.77	0.76
s+e+c+b	0.85	0.78	0.81	0.77

The test consisted of a 5-fold cross validation using dataset B100 where interface residues were defined using DIMPLOT (Wallace, et al., 1995). The first column indicates the combination of features used: structural (s), energy (e), conservation (c), and B-factors (b). The second, third, fourth, fifth columns contain AUC values for Voronoi Diagram, sphere (15 Å cut-off), 9-residue sliding window, and single residue (no environment), respectively.

ISSN: 1471-2105