Accurate and Efficient Corrections for Missing Dispersion Interactions in Molecular Simulations

In simulations, molecular dispersion interactions are frequently neglected beyond a cutoff of around 1 nm. In
some cases, analytical corrections appropriate for isotropic systems are applied to the pressure and/or the
potential energy. Here, we show that in systems containing macromolecules, either of these approaches introduce
statistically significant errors in some observed properties; for example, the choice of cutoff can affect computed
free energies of ligand binding to proteins by 1 to 2 kcal/mol. We review current methods for eliminating this
cutoff-dependent behavior of the dispersion energy and identify some situations where they fail. We introduce
two new formalisms, appropriate for binding free energy calculations, which overcome these failings, requiring
minimal computational effort beyond the time required to run the original simulation. When these cutoff
approximations are applied, which can be done after all simulations are completed, results are consistent
across simulations run with different cutoffs. In many situations, simulations can be run with even shorter
cutoffs than typically used, resulting in increased computational efficiency.