Protein structure prediction using an associated memory Hamiltonian and all-atom molecular dynamics simulations

Abstract

We tried to predict the tertiary structure of the 63-residue-long alpha-helical protein, 1r69, from the amino acid sequence with the assumption that the locations of α-helical residues are known. We applied two approaches. One approach is to implement all-atom molecular dynamics (MD) simulations of segments of the target protein and use the snapshot structures of these simulations as memory sets in the associated memory Hamiltonian, which uses coarse-grained model of protein structure and describes the effect of solvent by a water-mediated long range interaction potential. The other approach is to implement all-atom MD simulations with implicit water model applying additional biasing potential functions to reduce the radius of gyration and induce the formation of secondary structures for the helical residues. In the coarse grained model of the associated memory Hamiltonian we tried two different sets of memory to see the effect of the local structural signals in the memory set. We found that the predicted results strongly depend on the structures used in the memory set. The predicted results from the associated memory Hamiltonian give a structure with RMSD value of 1.977 Å with respect to the native structure. The predicted results from the biased all atom MD simulation method give a structure with RMSD value of 2.971 Å.