Population Based Reweighting of Scaled Molecular Dynamics

• Published in: The journal of physical chemistry. B Vol. 117; no. 42; pp. 12759 - 12768
• Main Authors: Sinko, William, Miao, Yinglong, de Oliveira, César Augusto F, McCammon, J. Andrew
• Format: Journal Article
• Language: English
• Published: Washington, DC American Chemical Society 24.10.2013
• Subjects: Alanine - chemistry; Biological and medical sciences; Biomolecules; Computer simulation; Conformational dynamics in molecular biology; Dipeptides - chemistry; Dipeptides - metabolism; Free energy; Fundamental and applied biological sciences. Psychology; Molecular biophysics; Molecular dynamics; Molecular Dynamics Simulation; Oligopeptides - chemistry; Oligopeptides - metabolism; Potential energy; Protein Structure, Secondary; Sampling; Sampling methods; Statistics; Thermodynamics; Potential energy surfaces; Computing method; Free energy; Protein; Molecular dynamics method; Conformation
• ISSN: 1520-6106; 1520-5207
• DOI: 10.1021/jp401587e

Molecular dynamics simulation using enhanced sampling methods is one of the powerful computational tools used to explore protein conformations and free energy landscapes. Enhanced sampling methods often employ either an increase in temperature or a flattening of the potential energy surface to rapidly sample phase space, and a corresponding reweighting algorithm is used to recover the Boltzmann statistics. However, potential energies of complex biomolecules usually involve large fluctuations on a magnitude of hundreds of kcal/mol despite minimal structural changes during simulation. This leads to noisy reweighting statistics and complicates the obtainment of accurate final results. To overcome this common issue in enhanced conformational sampling, we propose a scaled molecular dynamics method, which modifies the biomolecular potential energy surface and employs a reweighting scheme based on configurational populations. Statistical mechanical theory is applied to derive the reweighting formula, and the canonical ensemble of simulated structures is recovered accordingly. Test simulations on alanine dipeptide and the fast folding polypeptide Chignolin exhibit sufficiently enhanced conformational sampling and accurate recovery of free energy surfaces and thermodynamic properties. The results are comparable to long conventional molecular dynamics simulations and exhibit better recovery of canonical statistics over methods which employ a potential energy term in reweighting.