The use of momentum-space descriptors for predicting octanol-water partition coefficients

We extend previous work on quantitative structure-activity and structure-property relationships using molecular descriptors based on quantities determined from the momentum-space (p-space) electron density. In particular, we introduce new molecular descriptors that are related to the p-space information entropy. We also examine the use of a simple molecular shape descriptor, X. For the LC50 toxicity data of a series of saturated alcohols, the simple shape descriptor is found to be remarkably successful, and a correlation is observed between X and the entropy-like p-space descriptors. We develop a promising 13-descriptor regression model for the log P values of a set of 76 chemically diverse molecules. Similar approaches, combining simple classical parameters with p-space quantities, are likely to prove useful for more complex QSAR/QSPR problems. (c) 2005 Elsevier B.V. All rights reserved.