The accurate prediction of a ligand’s binding mode to its target receptor is central to structure-based drug design. Docking is commonly used to predict the 3D structure of ligand:protein complexes based on the structures of the individual components. The docking of flexible peptide ligands to their receptors, however, is non-trivial due to the complex search space resulting from the very large number of potential ligand-protein interactions. One way of reducing the complexity of the search space is to guide the docking process with restraints based on experimentally-derived information regrading the interacting residues at the peptide-receptor interface. The question is, how many and what type of restraints (and thus what type of experimental information) is most effective for confidently predicting the structure of the ligand-protein complex.
To address this problem we used the co-crystal structure of the complex formed between the 40-residue peptide PcTx1 and acid sensing ion channel (ASIC) 1a as a case study. A series of HADDOCK1 docking experiments were carried out using a combination of ambiguous interaction restraints (AIRs) (e.g., from alanine scanning mutagenesis or NMR chemical shift mapping experiments) and unambiguous distance restraints (UDRs) (e.g. from function recovery double mutants) in order to determine which combination of restraints is most effective in reproducing the binding mode observed in the co-crystal structure. Using only information regarding the ligand as AIR, even if the approximate location of the binding site of the protein is known, was insufficient to correctly predict the binding mode in our case. It was found that at least 4 UDR, were required to predict the binding mode with a 50% confidence. Adding information on pharmacophore residues on the toxin in the form of AIRs to the UDRs did not improve the predictive ability. The efficacy of different combinations of restraints will be presented.