. Author manuscript; available in PMC: 2012 Feb 28.

Published in final edited form as: J Chem Inf Model. 2011 Jan 7;51(2):214–228. doi: 10.1021/ci100257s

Table 4.

Correlation evaluation of the PHOENIX scoring function using different training sets on the PDBbind v2002 (n = 796) database. Correlation statistics include Pearson correlation coefficient (R_p), Spearman correlation coefficient (R_s), standard deviation (SD), mean error (ME), slope (a) in the linear regression (y = ax + b), and intercept (b). The number after the scoring function indicates the total number of complexes used for training (e.g., PHOENIX_68, training set of 68 complexes). The final PHOENIX scoring function (PHOENIX_112) is indicated in bold lettering.

Scoring Function	Rp	Rs	SD	ME	a	b
PHOENIX_68	0.499	0.518	2.04	1.62	0.33	4.09
PHOENIX_82	0.41	0.449	6.27	5.94	0.38	8.34
PHOENIX_105	0.473	0.502	2.06	1.6	0.43	3.25
PHOENIX_112	0.524	0.559	1.98	1.56	0.37	4.27
PHOENIX_127	0.517	0.534	2.07	1.65	0.33	4.77
PHOENIX_140	0.424	0.445	2.91	2.41	0.41	2.32
PHOENIX_153	0.333	0.339	2.24	1.81	0.21	4.54
PHOENIX_162	0.492	0.513	2.17	1.72	0.34	3.62