Proteins, nucleic acids, lipid molecules and other biomolecules play key roles in a cell’s biological functions. Their atomistic structures are usually determined by experiments, using crystallography, nuclear magnetic resonance (NMR), and cryo-electron microscopy. It is also important to determine their conformational dynamics, for understanding the structure-dynamics-function relationship will help advance the life sciences and new drug discoveries. Molecular dynamics (MD) simulations of biomolecules are often used for this purpose, but they need to become much faster to overcome the current limitations in spatial and temporal scales.
We are developing a new MD simulation program: GENESIS (Generalized-Ensemble Simulation System) and improving its performance through optimization and parallelization.In addition, we are developing novel computational algorithms for modeling largescale conformational changes of biomolecules, such as the replica-exchange method and the string method that connect between two structures. Multi-scale modeling techniques combining atomistic models, coarse-grained models, and hybrid quantum mechanics/molecular mechanics (QM/MM) models are also necessary for our simulations.
Coupling between simulations and experiments is yet another important subject in computational biology. We are applying several data-analysis methods, data assimilation and machine learning to link single-molecule-measurement time-series with MD simulations of biomolecules.
The first molecular dynamics simulations of an atomistic model of the bacteria cytoplasm
Conformational dynamics of a protein or nucleic acid is often investigated with conventional MD simulations. In such an analysis, a single biomolecule is solvated in solvent water and is simulated as long as possible to examine the structure-dynamics-function relationship. As well as molecular interaction between biomolecules in cellular environments such as the cytoplasm, investigation of the biological membrane and cellular nuclei are also essential for understanding biological cellular functions. However, due to the limitations in computational resources and power of MD software, these studies have not been performed until now.
We have optimized and parallelized our MD software GENESIS for simulating large biological systems that contain more than 10 million atoms. Mycoplasma genitalium is one of the smallest bacteria whose genetic and proteomic information is abundant. Based on this information, an atomistic model of the bacteria cytoplasm with its many proteins, RNAs, ribosomes, ions, metabolites like ATP, and water molecules was built, and was then simulated using GENESIS on the K computer. The detailed trajectory analysis produced new insights into protein stability, diffusion, and specific and nonspecific interactions in the cytoplasmic environments. Biomolecular simulations in cellular environments are expected to contribute not only to the life sciences but also to new drug discoveries.
Computer simulation of an atomistic model of the bacteria cytoplasm