ABSTRACTPredicting crystallographic B-factors of a protein from a conventional molecular dynamics simulation is challenging in part because the B-factors calculated through sampling the atomic positional fluctuations in a picosecond molecular dynamics simulation are unreliable and the sampling of a longer simulation yields overly large root mean square deviations between calculated and experimental B-factors. This article reports improved B-factor prediction achieved by sampling the atomic positional fluctuations in multiple picosecond molecular dynamics simulations that use uniformly increased atomic masses by 100-fold to increase time resolution. Using the third immunoglobulin-binding domain of protein G, bovine pancreatic trypsin inhibitor, ubiquitin, and lysozyme as model systems, the B-factor root mean square deviations (mean ± standard error) of these proteins were 3.1 ± 0.2–9 ± 1 Å2for Cα and 7.3 ± 0.9–9.6 ± 0.2 Å2for Cγ, when the sampling was done, for each of these proteins, over 20 distinct, independent, and 50-picosecond high-mass molecular dynamics simulations using AMBER forcefield FF12MC or FF14SB. These results suggest that sampling the atomic positional fluctuations in multiple picosecond high-mass molecular dynamics simulations may be conducive toa prioriprediction of crystallographic B-factors of a folded globular protein.
Reactive molecular dynamics simulation of the high-temperature pyrolysis of 2,2′,2′′,4,4′,4′′,6,6′,6′′-nonanitro-1,1′:3′,1′′-terphenyl (NONA)
