Identifying Potent Compounds Using Pairwise Consensus Methods

Molecular docking is a widely used method within the in silico compound screening process of modern drug discovery. The accuracy of this method for predicting high-affinity small-molecule binders for a target protein from a large chemical library can be substantially improved by combining individual docking tools for cross-validation. This traditional consensus strategy typically relies on averaging scores or ranks obtained from molecular docking, which are, however, vulnerable to false positives and thus exploit shortcomings from scoring functions. To overcome this remarkable weakness, we developed here the pairwise consensus score (PCS) algorithm. PCS integrates structural similarity information on ligand-receptor complexes to evaluate predicted conformations and penalize highly dissimilar docked poses. To demonstrate the versatility of PCS, we developed a consensus docking protocol for targeting G protein-coupled receptors (GPCRs) that are among the most important targets for modern drug discovery. In particular, we screened a large compound library for highly potent antagonism ligands to an important GPCR therapeutic target, the neurokinin 1 receptor, and found several compounds targeting the receptor with ten-picomolar activity. Notably, these highly active compounds show a totally different chemical structure from that of previously reported NK1 binders. This opens exciting opportunities to develop drugs with unique alternative pharmacological features and therapeutic value.