QSAR study on hERG inhibitory effect of kappa opioid receptor antagonists by linear and non-linear methods

TLDR: In this article, a dataset containing 45 inhibitors of the human ether-a-go-go -go voltage-gated ion-channel with known inhibitory was used, where the whole dataset was divided into a training set and a test set based on K-means clustering technique.

Abstract: The kappa opioid receptor antagonists have been studied for their quantitative structure–activity relationships. A dataset containing 45 inhibitors of the human ether-a-go–go voltage-gated ion-channel with known inhibitory was used. The whole dataset was divided into a training set and a test set based on of K-means clustering technique. Multiple linear regressions were employed to model the relationships between molecular descriptors and biologic activity of molecules using stepwise and genetic algorithm methods as variable selection tools. A comparison between the attained results indicated the superiority of the genetic algorithm over the stepwise multiple regression method in the feature selection. Support vector machine was also employed to model the non-linear structure–activity relationships. The models were validated using leave-one-out cross-validation, Y-randomization test, and applicability domain. The results showed that the linear model does not perform as well as the non-linear model in terms of predictive ability. The results suggest that the shape, relative negative charge, atomic masses, atomic polarizability, and atomic electronegativity are the main independent factors contributing to the hard inhibitory of the studied compounds.