Comment to: “Martini straight: Boosting performance using a shorter cutoff and GPUs” by D.H. de Jong, S. Baoukina, H.I. Ingólfsson, and S.J. Marrink

Abstract

In a recent study published in this journal, de Jong et al. investigated the efficiency improvement reached thanks to new parameter sets for molecular dynamics simulations using the coarse-grained Martini force-field and its implementation in the Gromacs simulation package (de Jong et al., 2016). The advantages of the new sets are the computational efficiency and the conservation of the equilibrium properties of the Martini model. This article reports additional tests on the total energy conservation for zwitterionic lipid bilayer membranes. The results show that the conclusion by de Jong et al. on the total energy conservation of the new parameter sets, based on short simulations and homogeneous systems, is not generalizable to long lipid bilayer simulations. The energy conservation of the three parameter sets compared in their article (common, new and new-RF) differ if one analyzes sufficiently long trajectories or if one measures the total energy drifts. In practice, when total energy conservation is important for a Martini lipid bilayer simulation, we would consider either keeping the common set, or carefully testing the new-RF set for energy leaks or sources before production use.