3D QSAR investigations on antimalarial naphthylisoquinoline alkaloids by comparative molecular similarity indices analysis (CoMSIA), based on different alignment approaches.

3D QSAR models based on the CoMSIA descriptor fields were established using a diverse data set of 53 antimalarial biaryl compounds (tested in vitro against a chloroquine-resistant strain of Plasmodium falciparum), consisting mainly of naphthylisoquinoline alkaloids, but also including phenylanthraquinone structures and naphthylindenes. For the alignment, two commercially available automated approaches, FLEXS and GASP, were compared; initially none of them succeeded in treating the important phenomenon of axial chirality correctly, but after some manual refinement of the alignments initially obtained, the best overall model, based on a modified FLEXS alignment, showed a q(2) (cross-validated r(2)) of 0.818 (eight components), using only the hydrophobic and the H-bond donor and acceptor fields. Using a test set of five compounds the model showed a squared multiple correlation coefficient for the test set (predictive r(2)) of 0.578. The analysis of the 3D contour maps permitted interesting conclusions about the effects of particular functional groups on the biological activity and will now guide the design of novel, hopefully even more active compounds.