Fast Prediction Of Biomolecular Conformational Equilibria Using One-step Perturbation

Computer simulation using molecular dynamics are increasingly used in biomolecular simulations. Molecular mechanics force fields offer empirical but computationally efficient approximations to quantum-mechanical models. Yet, the quality of the obtained results depends largely on the quality of the force field used. The one-step free energy perturbation approach can be applied to obtain conformational state-specific free energy differences associated with changes in force field parameters, and thus offers the possibility to consider conformational equilibria during force-field parameterization. Using the alanine deca-peptide in explicit water solution as a model, theα-helical andβ-hairpin state-specific free energy differences associated with force field changes between two widely used parameter sets of the GROMOS force field were determined using one-step perturbation. The results mostly deviated by only 1 kJ mol^-1 from thermodynamic integration results, suggesting that the convergence ranges of one-step perturbation were large enough to cover the substantial changes in non-bonded parameters between the two parameter sets. It was also found that one-step perturbation may give larger errors when the changes from the reference state include a large decrease in van der Waals radius. Such kind of perturbations should be avoided when applying this method. According to the free energy results, theα-helical state of the alanine deca-peptide is destabilized by 15 kJ mol^-1 relative to theβ-hairpin state when going from GROMOS 43A1 to 53A6. By applying one-step perturbation to analyze the effects of perturbing individual parameters, the differential stabilization of the two secondary structure states can be traced to the changes in van der Waals parameters, especially a van der Waals parameter involved in third-neighbor interactions. One-step perturbation was furthur used to correctly predict many folding equilibria of a poly-β-peptide with different side-chain substitutions from just one single MD simulation using an unphysical reference state. The evaluated methodology constitutes a powerful molecular simulation methodology by reducing the number of required separate simulations by orders of magnitude.