Assessing the Ability of Molecular Dynamics Force Fields to Capture Conformational Dynamics of Amino Acid Residues in Water*

Molecular dynamics (MD) is a powerful tool for studying intrinsically disordered proteins, however, its reliability depends on the accuracy of the force field.  We here assess Amber ff19SB & ff14SB, OPLS-AA/M, and CHARMM36m with respect to their capacity to capture intrinsic conformational dynamics of a guest amino acid residue x of unblocked GxG in water. Using published spectroscopic data (5 J-coupling constants and amide I' band profiles), MD force fields are evaluated with respect to their ability to reproduce experimental data of 14 residues x (x= G, A, L, V, I, F, Y, D^P, E^P, R, C, N, S, T) in GxG peptides through reduced χ² functions.  As a benchmark, a Ramachandran distribution is created as a linear combination of Gaussian sub distributions to best fit the experimental data.  Our results show that the Gaussian model outperforms all four MD force fields for all 14 guest residues. The major weaknesses of the four MD force fields include insufficient variability of the polyproline II (pPII) population among the guest residues, oversampling of antiparallel β-strand at the expense of transitional β-strand region, inadequate sampling of turn-forming conformations for ionizable and polar residues, and lack of guest residue-specificity in the Ramachandran distributions.  While Amber ff19SB performs worse than the other three force fields upon comparison of χ² values, it best accounts for guest residue-specificity, particularly pPII variability, among residues compared to the other three force fields.