Substituting random forest for multiple linear regression improves binding affinity prediction of scoring functions: Cyscore as a case study

Table 4 Cross validation results of the four models on the five partitions of PDBbind v2013 refined set (N = 2959) in terms of root mean square error RMSE, standard deviation SD in linear correlation, Pearson correlation coefficient Rp and Spearman correlation coefficient Rs

				MLR::Cyscore				RF::Cyscore				RF::CyscoreVina				RF::CyscoreVinaElem
#	N	RMSE	SD	Rp	Rs	RMSE	SD	Rp	Rs	RMSE	SD	Rp	Rs	RMSE	SD	Rp	Rs
1	592	1.66	1.66	0.560	0.555	1.60	1.60	0.601	0.588	1.41	1.41	0.708	0.709	1.33	1.33	0.748	0.746
2	592	1.62	1.62	0.589	0.600	1.51	1.51	0.657	0.641	1.38	1.37	0.730	0.725	1.30	1.29	0.764	0.766
3	592	1.69	1.70	0.531	0.529	1.66	1.66	0.561	0.545	1.49	1.49	0.668	0.665	1.41	1.41	0.711	0.709
4	592	1.68	1.68	0.542	0.557	1.63	1.63	0.580	0.576	1.51	1.51	0.657	0.661	1.41	1.41	0.711	0.722
5	591	1.65	1.65	0.559	0.553	1.57	1.57	0.615	0.586	1.42	1.42	0.701	0.692	1.30	1.30	0.758	0.749
avg		1.66	1.66	0.556	0.559	1.59	1.59	0.603	0.587	1.44	1.44	0.693	0.690	1.35	1.35	0.738	0.738

ISSN: 1471-2105