Table 5.

Comparison of cross-predictions between the Ki and IC50 models and datasets

	BDB Ki training set	BDB Ki test set	BDB IC50 training set	BDB IC50 test set
number of compounds	225	56	1319	322
Ki ALL model	0.99	0.77	0.64	0.69
IC50 ALL model	0.64	0.49	1.00	0.87

Note: Pearson correlations between experimental affinities and the random forest predictions are reported.