QSAR modeling for predicting carcinogenic potency of nitroso-compounds using 0D-2D molecular descriptors

Abstract

This paper reports a QSAR study for predicting carcinogenic potency of nitroso-compounds bioassayed in female rats administrated by gavage as oral route. Several different theoretical molecular descriptors, - 0D, 1D and 2D - calculated only on the basis of knowledge of the molecular structure and an efficient variable selection procedure, such as Genetic Algorithm, led to models with satisfactory predictive ability. But the best-final QSAR model is based on the combination between; 0D, 1D and 2D-DRAGON descriptors capturing a reasonable interpretation. This QSAR model is able to explain around 86% of the variance in the experimental activity and manifest good predictive ability as indicated by the higher q2s of cross validations, which demonstrate the practical value of the final QSAR model for screening and priority testing. This model can be applied to nitroso-compounds different from the studied nitroso-compounds (even those not yet synthesized) as it is based on theoretical molecular descriptors that might be easily and rapidly calculated.