A novel approach was developed to rationally interface structure- and ligand-based drug design through the rescoring of docking poses and automated generation of molecular alignments for 3D quantitative structure-activity relationship investigations. The procedure was driven by a genetic algorithm optimizing the value of a novel fitness function, accounting simultaneously for best regressions among binding-energy docking scores and affinities and for minimal geometric deviations from properly established crystal-based binding geometry. The GRID/CPCA method, as implemented in GOLPE, was used to feature molecular determinants of ligand binding affinity for each molecular alignment. In addition, unlike standard procedures, a novel multipoint equation was adopted to predict the binding affinity of ligands in the prediction set. Selectivity was investigated through square plots reporting experimental versus recalculated binding affinities on the targets under examination. The application of our approach to the modeling of affinity data of a large series of 3-amidinophenylalanine inhibitors of thrombin, trypsin, and factor Xa generated easily interpretable and independent models with robust statistics. As a further validation study, our approach was successfully applied to a series of 3,4,7-substituted coumarins, acting as selective MAO-B inhibitors.