Nonequilibrium, multiple-timescale simulations of ligand-receptor interactions in structured protein systems

Predicting the long-time, nonequilibrium dynamics of receptor-ligand interactions for structured proteins in a host fluid is a formidable task, but of great importance to predicting and analyzing cell-signaling processes and small molecule drug efficacies. Such processes take place on timescales on the order of milliseconds to seconds, so "brute-force" real-time, molecular or atomic simulations to determine absolute ligand-binding rates to receptor targets and over a statistical ensemble of systems are not currently feasible. In the current study, we implement on real protein systems a previously developed 3-5 hybrid molecular dynamics/Brownian dynamics algorithm, which takes advantage of the underlying, disparate timescales involved and overcomes the limitations of brute-force approaches. The algorithm is based on a multiple timescale analysis of the total system Hamiltonian, including all atomic and molecular structure information for the system: water, ligand, and receptor. In general, the method can account for the complex hydrodynamic, translational-orientational diffusion aspects of ligand-docking dynamics as well as predict the actual or absolute rates of ligand binding. To test some of the underlying features of the method, simulations were conducted here for an artificially constructed spherical protein "made" from the real protein insulin. Excellent comparisons of simulation calculations of the so-called grand particle friction tensor to analytical values were obtained for this system when protein charge effects were neglected. When protein charges were included, we found anomalous results caused by the alteration of the spatial, microscopic structure of water proximal to the protein surface. Protein charge effects were found to be highly significant and consistent with the recent hypothesis of Hoppert and Mayer (Am Sci 1999;87:518-525) for charged macromolecules in water, which involves the formation of a "water dense region" proximal to the charged protein surface followed by a "dilute water region." We further studied the algorithm on a D-peptide/HIV capside protein system and demonstrated the algorithms utility to study the nonequilibrium docking dynamics in this contemporary problem. In general, protein charge effects, which alter water structural properties in an anomalous fashion proximal to the protein surface, were found to be much more important than the so-called hydrodynamic interaction effects between ligand and receptor. The diminished role of hydrodynamic interactions in protein systems allows for a much simpler overall dynamic algorithm for the nonequilibrium protein-docking process. Further studies are now underway to critically examine this simpler overall algorithm in analyzing the nonequilibrium protein-docking problem.