Table 3.

Exploration of two collections of empirical descriptors for the quantitative structure–activity relationship (QSAR) Random Forests (RF) model of pMIC against MRSA for the training set with an OOB estimation.

Descriptors (#)	R2	RMSE 1	MAE 2	% Error ≥ 0.5/% Error < 0.5 3
MACCS (166) 4	0.555	0.649	0.477	37/63
Sub (307) 4	0.509	0.681	0.505	39/61
SubC (307) 4	0.563	0.642	0.468	36/64
PubChem (881) 4	0.561	0.643	0.467	36/64
CDK (1024) 4	0.551	0.652	0.471	36/64
CDK Ext (1024) 4	0.572	0.636	0.456	34/66
1D2D (218) 5	0.364	0.775	0.600	49/51

¹ Root mean squared error. ² Mean absolute error. ³ Percent of molecules predicted with absolute error above or below 0.5. ⁴ Fingerprints. ⁵ Molecular descriptors.