P2Y1 is a G protein-coupled receptor (GPCR) activated by ADP. Its antagonists impede platelet aggregation in vivo and are potential antithrombotic agents. How can we exploit homology models for a quantitative prediction of the biological activities? When a training set is available, combining structure-based and ligand-based models may be the answer. Using a set of antagonists previously synthesized and tested in our laboratories, we tested techniques based on the estimation of the receptor-ligand interactions, such as docking and Linear Interaction Energy (LIE), and QSAR techniques based on the 2D and 3D properties of the ligands. The structural alignment and the conformation of the molecules were derived from docking. To benefit from the strength of each technique and compensate their inherent limits, we correlated by PLS the predictions of the individual methodologies to the experimental potencies, generating a consensus model which outperformed the internal and external predictivity of any individual model.