Prediction of the complexation of structurally diverse compounds with β-cyclodextrin using structural descriptors derived from electrostatic potentials on molecular surface and different chemometric methods

Abstract

A quantitative structure–property relationship (QSPR) study was performed for predicting the complexation of structurally diverse compounds with β-cyclodextrin (β-CD). Six statistical methods, which include conventional multiple linear regression (MLR) and partial least-squares regression (PLS), and some up-to-date modeling techniques—support vector machine (SVM), least-squares support vector machine (LSSVM), random forest (RF) and Gaussian process (GP), were utilized to build the QSPR models. Systematical validations including internal leave-one-out cross-validation, the validation for external test set, as well as a more rigorous Monte Carlo cross-validation were also performed to confirm the reliability of the constructed models. Among these modeling methods, the GP, which can handle linear and nonlinear-hybrid relationship through a mixed covariance function, showed the best fitting and predictive abilities. The coefficient of determination rpred2 and root mean square error of prediction (RMSEP) for the external test set were 0.832 and 0.373, respectively. Physical meanings of all structural descriptors introduced, which include six quantities derived from electrostatic potential on molecular surface (ESPMS) and the energy level of highest occupied molecular orbital (EHOMO), were elucidated. Some simple comparisons with previous QSPR results for the same or similar data sets were also made.