By Stefan Doerr
Molecular dynamics simulations (MD) can provide an atomic level resolution of biological processes at very high temporal resolution, but it comes with its own set of limitations; the most pronounced ones being the accuracy of the forcefields and the time sampling limitations.
Yet, we believe that there are further important problems: the data analysis and the reproducibility of experiments. In the last few years, specialized hardware, high-throughput methods and advanced sampling techniques have come to significantly improve molecular dynamics, allowing them to reach aggregate simulation times of multiple milliseconds. Forcefields have also dramatically improved. This increase of simulation accuracy and data has led to the necessity of a more standardized methodology for preparing, executing and handling thousands of individual trajectories.
Recently we have been working on provide an environment for molecular discovery which simplifies every step of molecular modelling and simulations. We call this environment high-throughput molecular dynamics (HTMD) and the final paper is now published in JCTC. A short intro is provided in this blog.
Investigating biological processes using MD usually requires the processing of large amounts of data and files, using various tools and adapting to peculiarities of many different software packages developed over several decades. With all these fragile set of tools, it is hard to follow the steps of a workflow that lead from the original PDB to the results, even for the scientist who wrote the workflow. Secondly, it is hard to extend the functionality of the tools because of such diversity of languages and the absence of a common programming environment where to introduce new extensions.
HTMD is our vision of a unified platform, a programmable workspace for simulation based molecular discovery. We name it HTMD (high-throughput molecular dynamics) to indicate the fact that it allows the handling of thousands of simulations and multiple systems in a controlled manner. HTMD extends the Python programming language with functions and classes to handle molecular systems at different levels while abstracting implementation details and best-practice knowledge. Python is a scripting language which enjoys widespread usage in the scientific community and thus provides an ideal platform on which to develop and distribute HTMD. HTMD’s functionalities span from molecular structure manipulation to visualization, to preparing and executing molecular simulations on different compute resources and data analysis, including the use of Markov state models (MSMs) to identify slow events, kinetic rates, affinities and pathways.
Stefan Doerr, Matthew J. Harvey, Frank Noé, and Gianni De Fabritiis, HTMD: High-throughput molecular dynamics for molecular discovery, in press JCTC 2016