High-dimensional QSAR prediction of anticancer potency of imidazo[4,5-b]pyridine derivatives using adjusted adaptive LASSO

Abstract

In high-dimensional quantitative structure-activity relationship (QSAR) studies, identifying relevant molecular descriptors is a major goal. In this study, a proposed penalized method is used as a tool for molecular descriptors selection. The method, called adjusted adaptive least absolute shrinkage and selection operator (LASSO) (AALASSO), is employed to study the high-dimensional QSAR prediction of the anticancer potency of a series of imidazo[4,5-b]pyridine derivatives. This proposed penalized method can perform consistency selection and deal with grouping effects simultaneously. Compared with other commonly used penalized methods, such as LASSO and adaptive LASSO with different initial weights, the results show that AALASSO obtains the best predictive ability not only by consistency selection but also by encouraging grouping effects in selecting more correlated molecular descriptors. Hence, we conclude that AALASSO is a reliable penalized method in the field of high-dimensional QSAR studies