By analyzing the structural characteristics of the studied compounds encoded in the afore-mentioned descriptors along with their effects on pIC50 inhibitory activity, we have been able to design eleven new chemical inhibitors. Herein, both MLR and MNLR models which exhibited excellent performance and met OECD criteria were exploited to predict inhibitory activities. Likewise, the nonlinear model's predictive power has been measured internally through the Loo/cross-validation procedure with coefficient R !'($%%) ' and externally through test set compounds with external prediction coefficient R ()*+ '. The Loo/cross-validation procedure, the Y-scrambling test, Golbraikh-Tropsha's criteria and the applicability area have all been utilized to evaluate the linear model's performance accuracy. Among the 16 quantum and physicochemical descriptors measured, only four optimal molecular descriptors have been employed to perform QSAR models, i.e., density, number of H-bond acceptors, octanol/water partition coefficient, and LUMO energy. To accelerate this process, linear and non-linear models of thirty-nine pyrimidine-2,4-dione derivatives have been constructed by exploiting PCA, MLR, and MNLR statistical techniques available in the XLSTAT software, as well as the (DFT/ Beck3LYP/6 − 31G (d, p)) approach.

The main target of present QSAR modeling is to pave the way for the development of new pyrimidine-2,4-dione derivatives and predict their HIV reverse transcriptase-associated RNase H inhibitory activity.

Xlstat 3dplot