Abstract

In silico molecular models of receptor-ligand complexes, built using molecular docking methods, are valuable as they potentially reveal the chemical interactions responsible for specific affinities. When applied to series of congeneric compounds, they help formulate theories about the effects of different substituents on affinity differences. Molecular docking provides (i) a pose where chemical interactions are optimized and (ii) an energy value indicating how favorable the interaction is. The capability of molecular docking for the first purpose is recognized, but it fails considerably in the second. It is widely known that energy values obtained by molecular docking are unreliable, which makes their application to congeneric series unable to correlate these computationally calculated energy values with laboratory-derived biological activities. Theoretically, an improved correlation could be obtained when protein flexibility is considered in the docking calculation; i.e., flexibility in the protein residues at the binding site can give access to more representative docking solutions. With this in mind, in this work, we present the novel web server CorrEA with a simple and innovative way of considering the flexibility of ligand-protein systems. To apply the method, users must generate a set of receptor conformations exhibiting significant variability within the binding site. Subsequently, they should cross-dock the ligand series they intend to study, to obtain various poses for each ligand across the different receptor conformations. CorrEA performs a genetic algorithm (GA) selection to extract a representative complex for each ligand that better adjusts the global correlation between calculated docking energy values and experimental logarithmic biological activities. In the end, CorrEA provides the ligand-protein pairs that produce the highest correlations. The new method was tested in several different cases to demonstrate its usefulness.