Proteochemometric modelling coupled to in silico target prediction: an integrated approach for the simultaneous prediction of polypharmacology and binding affinity/potency of small molecules

Table 1 PCM, Family QSAR and Family QSAM performance on the PCM dataset

	R ² _CV	RMSE _CV	R ² _{0 ext}	RMSE _ext
GBM PCM	0.75	0.64	0.79	0.59
GP PCM	0.75	0.65	0.76	0.63
RF PCM	0.74	0.66	0.77	0.62
SVM PCM	0.76	0.63	0.77	0.62
Family QSAM	0.07	1.24	0.09	1.22
Family QSAR	0.61	0.80	0.63	0.78
Inductive Transfer	0.72	0.68	0.76	0.63

Abbreviations: QSAM Quantitative Structure-Activity Modelling, QSAR Quantitative Structure-Activity Relationship, GBM Gradient Boosting Machine, GP Gaussian Process, RF Random Forest, SVM Support Vector Machine.
PCM, with R² _{0 test} and RMSE_test values of 0.79 and 0.59 pIC₅₀ units, outperforms both Family QSAR, with R² _{0 test} and RMSE_test values of 0.63 and 0.78 pIC₅₀ units, respectively, and Family QSAM, with with R² _{0 test} and RMSE_test values of 0.09 and 1.22 pIC₅₀ units, respectively.

ISSN: 1758-2946