Evaluation of the CHARMM36m force field combined with the OPC water model for protein simulations

The accuracy of atomistic molecular dynamics (MD) simulations depends on the accuracy of the used force field. For proteins many force fields have been developed so far, each in combination with a particular water model. Typically, combining a protein force field with a water model for which is has not been developed is not expected to yield reliable results. However, recently the combination of the CHARMM36m force field with the OPC water model was shown to accurately estimate the compactness and secondary structure content of intrinsically disordered proteins. Whether CHARMM36m+OPC yields similar accuracy for globular proteins as well has, however, not yet been evaluated systematically. Here, we benchmark this combination on a set of six different globular proteins. To this end, we performed 50 x 1 μs MD simulations per protein using CHARMM36m+OPC and, for comparison, the same simulations using the well-established Amber99SB-ILDN+TIP4P force field. We compared the generated ensembles by means of RMSD, radius of gyration and secondary structure content and also compared the conformational dynamics. We found that CHARMM36m+OPC generates less compact ensembles and shows higher barriers for conformational transitions. Furthermore, we compare ensembles of both force fields to experimental crystal structures, B-factors, and NMR chemical shifts. Here, we found that both force fields compare equally well to experimental data, with CHARMM36m+OPC ensembles agreeing with chemical shifts slightly better, whereas Amber99SB-ILDN+TIP4P ensembles better agree with crystal structures. Our results show that CHARMM36m+OPC and Amber99SB-ILDN+TIP4P yield similar accuracy for globular proteins. Combined with earlier results for disordered proteins, these findings suggest that CHARMM36m+OPC should provide good accuracy for a broad range of disordered and folded proteins as well.