MuSTAR MD: multi-scale sampling using temperature accelerated and replica exchange molecular dynamics.

A new and efficient conformational sampling method, MuSTAR MD (Multi-scale Sampling using Temperature Accelerated and Replica exchange Molecular Dynamics), is proposed to calculate the free energy landscape on a space spanned by a set of collective variables. This method is an extension of temperature accelerated molecular dynamics and can also be considered as a variation of replica-exchange umbrella sampling. In the MuSTAR MD, each replica contains an all-atom fine-grained model, at least one coarse-grained model, and a model defined by the collective variables that interacts with the other models in the same replica through coupling energy terms. The coarse-grained model is introduced to drive efficient sampling of large conformational space and the fine-grained model can serve to conduct more accurate conformational sampling. The collective variable model serves not only to mediate the coarse- and fine-grained models, but also to enhance sampling efficiency by temperature acceleration. We have applied this method to Ala-dipeptide and examined the sampling efficiency of MuSTAR MD in the free energy landscape calculation compared to that for replica exchange molecular dynamics, replica exchange umbrella sampling, temperature accelerated molecular dynamics, and conventional MD. The results clearly indicate the advantage of sampling a relatively high energy conformational space, which is not sufficiently sampled with other methods. This feature is important in the investigation of transition pathways that go across energy barriers. MuSTAR MD was also applied to Met-enkephalin as a test case in which two Gō-like models were employed as the coarse-grained model.