QM/NN QSPR models with error estimation: Vapor pressure and LogP

QSPR models for logP and vapor pressures of organic compounds based on neur al net interpretation of descriptors derived from quantum mechanical (semie mpirical MO; AM1) calculations are presented. The models are cross-validate d by dividing the compound set into several equal portions and training sev eral individual multilayer feedforward neural nets (trained by the back-pro pagation of errors algorithm), each with a different portion as test set. T he results of these nets are combined to give a mean predicted property val ue and a standard deviation. The performance of two models, for logP and th e vapor pressure at room temperature, is analyzed, and the reliability of t he predictions is tested.