Complete sampling of an enzyme reaction pathway: a lesson from gas phase simulations
This work addresses the sampling issue commonly accompanying the simulation of chemical reactions. Very often the sampling is severely limited by complexity of the phase space, possibly leading to poorly converged or inaccurate free energy profiles. We explored the factors governing the completeness of reaction path sampling for the rate limiting step of phenylethylamine oxidation by lumiflavin in the gas phase, a reaction important for the pharmacology of the central nervous system. The simulations utilize the free energy perturbation sampling technique together with the empirical valence bond methodology for the free energy calculations. The simplicity of the system allows for the acquisition of fully converged free energy profiles, even for simulation free of restraints. The bottleneck for convergence is in the noticeably poorer sampling statistics in the transition state region, which is resolved by performing sufficiently long simulation to ensure reversibility of all processes accompanying the reaction. In the present case, convergence is attained in microseconds of simulation, but the required simulation time generally depends on the complexity of the potential energy surface pertinent to the reaction. Accordingly, the use of restraints reduces the complexity of the phase space, decreasing the required time by about an order of magnitude. In the case of elementary nucleophilic substitution with even simpler potential energy surface convergence is reached already at a timescale of few nanoseconds. For related biomolecular reactions embedded in an enzyme, significantly longer simulation times may be needed, rendering the sampling problem exceedingly difficult and representing a challenge for advanced sampling techniques. Accordingly, suggestions are given for optimal simulation of biomolecular reactions based on the presently employed techniques and under the aforementioned limitations.