Scale bridging in molecular simulation

Abstract

Abstract. Multiscale and multiphysics approaches have become an integral part of the molecular modeling and simulation toolbox and are used to attack various real-world problems that would be out of reach without these techniques. This special topics issue is devoted to a critical appraisal of some of the most popular scale bridging techniques for molecular simulation. It features regular articles and a "Discussion and Debate" section, in which experts in the field discuss specific articles and general aspects of scale bridging techniques.Recent years have seen a growing interest in multiscale and multiphysics approaches for molecular simulation that have become a part of many molecular dynamics simulation packages. On the one hand, the advent of such scale-bridging techniques has enabled to apply molecular dynamics simulation to real world problems that would have been out of reach otherwise [1]. On the other hand, the growing relevance of scale-bridging for computations has raised issues regarding the theoretical (mathematical and/or physical) foundation of the methods, the accuracy of the results, and the robustness of the algorithms, which play a key role for the predictive power of molecular simulations. For example, molecular drug design uses multiscale molecular dynamics simulations to study the flexibility of a protein and its interaction with a target molecule of interest; typical time scales range from femtoseconds (molecular vibrations) up to micro-or milliseconds (conformational changes) and length scales range from the size of an atom (a few angstroms) up to the size of a molecular complex with millions of atoms. As the interactions between ligand protein and target molecule are typically local, it is natural to resort to a coarse-grained description of the parts of the system that are far away from the interaction site. Coarse-grained models may use, e.g. beads to represent groups of atoms, and thus to reduce the number of unknowns in the simulation, which not only results in a speed up of the simulation, but also in a reduction of the amount of data that need to be analyzed (see Fig. 1). It is not always clear, however, whether the local dynamical interactions between protein and target molecule will lead to an overall (non-local) change in the protein backbone structure or just an unimportant side-chain rotation. Either way, the prediction of protein structures, binding free energies, or ligand association-dissociation rates require both model and algorithm to be reasonably accurate. For an efficient in silico a

Molecular Simulation for the Next Decade

“…Moreover the idea that the governing physical principles of simulated models should be framed, as much as possible, in a rigorous mathematical language is starting to become a standard procedure for many researchers, with the consequence that simulations are characterized by an increasing physical accuracy and reproducibility. [ 7 ] Such conceptual progress is taking place in parallel to what can be considered a true technical revolution in the field, the contribution of artificial intelligence. Machine learning is not only increasing in a substantial manner the numerical efficiency of simulation algorithms but is also contributing to the refinement of atomic and molecular models where the currently available ones do not offer the desired accuracy.…”