The key objectives of computational structure-based drug design include the prediction of the protein-ligand complex binding modes and estimation of the binding affinities. The overall affinity of a ligand for a receptor can be expressed as a balance between the strength of the interactions of a ligand to any particular binding-competent conformation of the receptor and the probability of occurrence of that conformation in the absence of the ligand. The receptor conformation probability distributions can be described by the free energy landscape of the receptor from which the strain free energy required to move from one conformation to another in the absence of a ligand may be estimated. The availability of large datasets of crystal structures in the PDB can provide information about the locations of free energy basins and their shapes. Here we utilize several methods in an effort to model the strain free energy of several receptors due to binding using the vast structural data publically available in the PDB. Clustering of 99 X-ray structures of HIV-1 reverse transcriptase at the flexible non-nucleoside inhibitor binding pocket elucidates eight discrete clusters, one of which displays a novel bound conformation of the functionally important primer grip. The clustering results served as a guide for replica exchange molecular dynamics simulations that offer a more in-depth look at the potential reorganization of the binding pocket. Clustering of 327 available Xray structures of HIV-1 protease reveals less discrete variability in the substrate envelope than HIV-1 reverse transcriptase but does reveal some receptor reorganization that may be due to a combination of mutations.

A linear response model for incorporation of receptor strain in modern protein-ligand binding affinity estimators is proposed. Receptor-receptor contact counts are employed as estimators for changes in receptor conformation due to binding of different ligands. Overall, the linear model produces apparent reduction in binding energy estimation errors and increases in the rank-order correlation with respect to initial values determined by the commercially available Glide 5.0 XP that does not take into account receptor reorganization. It also offers information as to the type of conformational changes, if any, that may contribute to the receptor reorganization energy. A null hypothesis test is constructed to evaluate the possibility of producing fits by chance alone. Finally, an alternative estimator approach using structurally significant intrareceptor distance descriptors, where there are less possible estimators, shows some promise for several drug targets. The model has the potential to allow for coarse-grained investigation of the conformational and energetic landscapes for binding inhibitors to flexible protein receptors.