Improved Statistical Sampling and Accuracy with Accelerated Molecular Dynamics on Rotatable Torsions.

Area of Science:

• Computational Chemistry
• Molecular Dynamics Simulations
• Biophysics

Background:

• Enhanced sampling techniques in molecular dynamics can suffer from reduced precision due to statistical weight variations and decreased effective sample size.
• Typical accelerated molecular dynamics (aMD) boosts all dihedrals, which can lead to inaccuracies.
• Selective acceleration of only rotatable dihedrals (RaMD) is proposed to mitigate these issues.

Purpose of the Study:

• To introduce and evaluate a novel accelerated molecular dynamics (aMD) approach, termed RaMD, which selectively boosts only rotatable dihedrals.
• To compare the performance of RaMD against normal molecular dynamics (MD) and aMD with all dihedrals boosted (all-aMD).
• To assess the impact of RaMD on the precision and effective sampling size in molecular simulations.

Main Methods:

• Development of the RaMD approach, focusing acceleration on rotatable dihedrals only.
• Benchmark simulations of two model dipeptides (Ace-Ala-Nme and Ace-Trp-Nme) using normal MD, all-aMD, and RaMD.
• Quantitative comparison of simulation performances using a theoretical framework to estimate effective sample size and uncertainty.

Main Results:

• RaMD significantly reduces the loss in effective sample size compared to all-aMD for equivalent acceleration levels and simulation lengths.
• RaMD achieves increased accuracy in sampling conformational space (e.g., φ-ψ space).
• The accuracy of RaMD is comparable to all-aMD but requires simulation lengths 5 to 1000 times shorter.

Conclusions:

• RaMD offers a significant improvement in both speed and accuracy over all-aMD for enhanced molecular dynamics sampling.
• By avoiding acceleration of non-essential dihedrals, RaMD prevents artificial increases in potential energy, enhancing simulation precision.
• RaMD presents a promising and highly efficient method for sampling larger and more complex biomolecules.