Peptide backbone sampling convergence with the adaptive biasing force algorithm.

Area of Science:

• Biomolecular simulations
• Computational chemistry
• Protein dynamics

Background:

• Complete Boltzmann sampling of reaction coordinates is difficult in unbiased molecular dynamics.
• Adaptive Biasing Force (ABF) is a method to enhance sampling by canceling average forces.
• ABF enables recovery of the unbiased Boltzmann distribution of states.

Purpose of the Study:

• To characterize the performance of the ABF algorithm for sampling Ramachandran φ/ψ dihedral angles.
• To evaluate ABF's effectiveness across all 20 amino acid dipeptides using molecular dynamics.
• To identify challenges and successes in applying ABF to biomolecular systems.

Main Methods:

• Employed the Adaptive Biasing Force (ABF) algorithm.
• Performed all-atom explicit-water molecular dynamics simulations.
• Applied ABF to the full Ramachandran φ/ψ backbone dihedral reaction coordinates for 20 amino acid dipeptides.

Main Results:

• Approximately 50% of dipeptides showed rapid and robust convergence of the potential of mean force.
• Branched-side chain amino acids (threonine, valine) and proline presented sampling challenges.
• Proline dipeptides exhibited trans-to-cis peptide bond isomerization, complicating sampling.

Conclusions:

• The ABF method is a robust approach for sampling the entire φ/ψ reaction coordinate for all 20 amino acids.
• ABF successfully samples high free-energy regions inaccessible to standard molecular dynamics.
• ABF significantly enhances the exploration of conformational space in biomolecular simulations.