Investigating the utility of momentum-space descriptors for predicting blood-brain barrier penetration

We investigate the possible use of families of momentum-space descriptors and of trivial classical descriptors for the prediction of blood-brain barrier penetration, expressed as log BB. A 12-descriptor model based on entropy-like momentum-space quantities and on the numbers of atoms of each type has good statistical quality for a set of 42 structurally diverse molecules. We also consider the inclusion in our models of some of the other descriptors that have been used in earlier models for these molecules. The resulting models are not expected to be useful as-is for making genuine predictions for much larger test sets, but the various results do demonstrate the potential benefits of incorporating momentum-space descriptors into QSAR models for predicting logBB.