Can molecular dynamics simulations provide high-resolution refinement of protein structure?

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Biophysics

Background:

• Protein structure prediction tools generate initial models that often require refinement.
• Generalized Born (GB) approximation offers efficient solvent treatment for simulations.
• Replica Exchange (REX) enhances sampling in molecular dynamics.

Purpose of the Study:

• To refine protein structures generated by prediction tools using molecular dynamics simulations.
• To incorporate structural information like secondary structures and topology into the refinement protocol.
• To assess the effectiveness of REX/GB simulations for protein structure refinement.

Main Methods:

• Molecular dynamics simulations employing the generalized Born (GB) approximation for solvent treatment.
• Incorporation of dihedral restraints for secondary structures and distance restraints for topology.
• Enhanced sampling using the replica exchange (REX) method.
• Clustering of simulation ensembles to identify refined models.

Main Results:

• Refinement simulations achieved significant improvements in C(alpha) RMSD for initial models close to the native state.
• REX/GB simulations of 3-5 nanoseconds were sufficient for convergence in small to medium single-domain proteins.
• Clustering yielded refined models with improved accuracy and better sidechain contacts and rotamer states.

Conclusions:

• Molecular dynamics simulations with REX/GB are effective for refining protein structures.
• The method shows limitations in refining long and flexible loops due to sampling and force field accuracy.
• Incorporating structural information and advanced sampling techniques enhances refinement accuracy.