In order to relate the geometric and electronic descriptors of the rutaecarpine analogous to their biological activity against cancer of the central nervous system (CNS), molecular quantum chemical calculations at B3LYP/6-31(d) level and statistical analysis were carried out for 21 rutaecarpine analogous. Out of the 86 molecular descriptors calculated, 5 were selected to build the principal component analysis (PCA) model. The PC1 component that accounts for 46.11% of the total variance of the data was alone able to discriminate completely the compounds into two classes: active and inactive. All molecular descriptors selected by PCA model are electronic parameters. The hierarchical cluster analysis (HCA) was also applied on the selected descriptors. Based on the 5 electronic descriptors selected, it is possible to suggest new compounds to be synthesized with activity against CNS cancer. In addition to that, a supervised partial least squares discriminant analysis (PLS-DA) model was built and successfully applied to discriminate rutaecarpine analogues, being validated through an independent test set and considered robust to overfitting.
central nervous system cancer; quinazoline-beta-carboline-5-one; B3LYP; rutaecarpine analogous
