A QSPR study of drug release from an arabinoxylan using ab initio optimization and neural networks

Abstract

A QSPR study on release of pharmacologically diverse drugs from a biocompatible matrix, arabinoxylan, by use of ab initio structure optimization and neural networks is reported. A total of 1685 quantum mechanical, physico-chemical and structural descriptors were calculated for 16 drug molecules. A heuristic approach combined with unsupervised forward selection was used to identify descriptors mechanistically related to response variables. The release models were developed using multiple linear regression (MLR) and artificial neural networks (ANN) and were validated by leave-one-out cross validation and y-scrambling techniques. The release was found to be controlled by softness, lipophilicity, unsaturation, atomic polarization, cyclic topology and geometry of the molecules. The quantitative–structure–property relationship (QSPR) models were found to be robust and highly predictive of release profile and mechanism of a drug molecule from the arabinoxylan matrix.