Receptor-based 3D-QSAR studies of checkpoint Wee1 kinase inhibitors.

One hundred and seventy-four pyrrolo[3,4-c]carbazole-1,3(2H,6H)-dione derivatives reported as inhibitors of the kinase Wee1 were used for a molecular docking and three-dimensional quantitative structure-activity relationship (3D-QSAR) study. Due to the availability of the three-dimensional structure of the Wee1 kinase a receptor-based alignment strategy was applied. Six available Wee1-inhibitor crystal structures were analyzed using the docking program GOLD resulting in a good reproduction of the experimentally derived position and interaction of the cocrystallized inhibitors. Since only a low correlation between docking scores and inhibitory activities was obtained for the series of 174 inhibitors a receptor-based 3D-QSAR study was performed, dividing the data set into 144 training set molecules and an external test set of 30 compounds. Besides the ligand alignment derived from the docking study we tested several other alignment procedures as basis for the 3D-QSAR analysis. The most predictive model was obtained using the alignment from the GOLD docking study. The CoMFA model was found to be robust (q(LOO)(2)=0.764 and r(2)=0.870). The predictive ability of the model was further examined by carrying out leave-20%-out and leave-50%-out cross-validation (q(2)=0.747 for leave-20%-out and 0.737 for leave-50%-out) and predicting the activities of 30 inhibitors used as external test set (r(pred)(2)=0.790). The graphical analysis of the CoMFA contour plot together with the key residues of the binding pocket provided important insight into the relevant interactions of the inhibitors. The results not only provide information about the essential features of potent Wee1 inhibitors but also show the advantage of using receptor-based alignment for 3D-QSAR analysis.