R-CCS Cafe 第238回 第1部

By performing molecular dynamics simulations of biomolecules on supercomputers like K or Fugaku, we can obtain big data of protein conformations in solution, membrane, or other cellular enviornments. Protein dynamics generally contains fast vibrational motions as well as slow and large conformational changes. Therefore, the analysis of molecular dynamics simulation trajectories is not so easy in most cases. Here, we introduce two different approaches to help protein dynamics analysis using molecular dynamics simulation trajectories. The first one is to reduce the number of particles after performing all-atom simulations. This approach was successfully applied to spike protein on the SARS-CoV-2. The second one is to reduce the dimension of conformational space with linear and non-linear dimensional reduction method. In particular, we recently proposed novel method based on UMAP and applied it to protein folding simulations. They would be useful for many biological simulations with appropriate approximations.