The influence of solid state information and descriptor selection on statistical models of temperature dependent aqueous solubility

Table 3 Top ranked results according to various scenarios for the temperature dependent solubility datasets

Dataset	CV protocol	Rank	Molecular descriptors	3D descriptors from crystal Structure?	Melting point descriptor included	Lattice energy descriptor included	Method	R²	RMSE (log units)
Klimenko_CS_False	rt	1st	IntegSub, SiRMSSub, Absolv, Ind, Rdk	False	True	False	RFR	0.92	0.70
Klimenko_CS_False	rt	2nd	Rdk, Absolv	False	True	False	RFR	0.92	0.70
Klimenko_CS_True	rt	1st	3D, IntegSub, SiRMSSub, Absolv, Ind, Rdk	False	True	False	RFR	0.85	0.83
Klimenko_CS_True	rt	2nd	Rdk, Absolv	False	True	False	RFR	0.85	0.83

All results were obtained without feature selection. All results are rounded to 2dp. The definitions of 2D molecular descriptors subsets are provided in Additional file 1: Table S1. All references to R² and RMSE denote arithmetic mean values obtained from cross-validation (CV = rt) and all model rankings were generated based on the mean RMSE values

ISSN: 1758-2946