Conformational sampling of protein flexibility in generalized coordinates: Application to ligand docking

Induced-fit effects are seldom considered in most docking programs, so protein mobility constitutes an obstacle in screening of chemical libraries against a single receptor conformation. To date, the best approach to incorporate protein backbone fluctuations in virtual screening is the use of multiple receptor conformations. Normal modes in internal generalized coordinates furnish an excellent way to represent receptor flexibility. Perturbation along these modes preserves covalent geometry and is especially suited for representing mid- and large-scale movements, since small changes in the internal variables might correspond to large changes in Cartesian space. However, the number of very-low-frequency normal modes thus calculated may be large. So we extended the definition of a measure of relevance of normal modes on a certain region of space. Working on a reduced spring model of protein kinase cAPK, we showed that only four modes have significant relevance on the gly-rich loop. Alternative receptor backbone conformations were generated by perturbing the structure along these relevant modes. Those conformations were complexed with non-native ligand balanol and subjected to global energy minimization to optimize side-chain positioning. The starting crystal structure and the one thus generated were used to perform a small-scale Receptor Ensemble Docking. It is shown that docking to the ensemble enhances both the RMSD values and enrichment factors. This modest benchmark indicates that normal modes-based deformation prediction along with ensemble docking could provide a solution to routinely account for receptor mobility in ligand docking.