Use of automatic relevance determination in QSAR studies using Bayesian neural networks

We describe the use of Bayesian regularized artificial neural networks (BRA NNs) coupled, with automatic relevance determination (ARD) in the developme nt of quantitative structure-activity relationship (QSAR) models. These BRA NN-ARD networks have the potential to solve a number of problems which,aris e in QSAR modeling such as the following: choice of model; robustness of mo del; choice of validation set; size of validation effort; and optimization of network architecture. The ARD method ensures that irrelevant or highly c orrelated indices used in the modeling are neglected as well as showing whi ch are the most important variables in modeling the activity data. The appl ication of the methods to QSAR of compounds active at the benzodiazepine an d muscarinic receptors as well as some, toxicological data of the effect of substituted benzenes on Tetetrahymena pyriformis is illustrated.