Table 2.

Comparison of novel stereoselective descriptor predictability with other published QSAR methods against the Cramer’s steroid set. Calculation of can be found in the methods section. Statistical model generation method is indicated as well as QSAR method employed are indicated for each reference.

QSAR Method	Model Creation	q2	Reference
Purely Spatial RDF-like stereochemistry	Artificial Neural Network	0.56
Property weight RDF-like stereochemistry (product)	Artificial Neural Network	0.74
Property weight RDF-like stereochemistry (sum)	Artificial Neural Network	0.78
Stochastic 3D-chiral linear indices	Multiple Linear Regression	0.87	[13]
Chiral Topological Indices	Stepwise Regression Analysis	0.85	[10]
Chiral Graph Kernels	Support Vector Machine	0.78	[11]
Chirality Correction and Topological Descriptors	K-nearest neighbor	0.83	[9]
Molecular Quantum Similarity Measures	Multilinear Regression	0.84	[24]
Shape and Electrostatic Similarity Matrixes	Non-linear Neural Network	0.94	[25]
Comparative Molecular Moment Analysis	Partial Least Squares (PLS)	0.83	[25]
Comparative Molecular Similarity Indices Analysis	PLS	0.67	[26]
Comparative Molecular Field Analysis	PLS	0.65	[20]
E-state Descriptors	PLS	0.62	[27]
Molecular Electronegativity Distance Vector	Genetic Algorithm PLS	0.78	[28]
Molecular Quantum Similarity Measures	Multilinear Regression and PLS	0.80	[29]