Bioexcel webinar #86 :MiMiC: A high-performance framework for multiscale MD simulations

These days, computational chemists can choose from a wide variety of software packages, each with its own unique features and strengths. Many of these programs also offer some support for modeling chemical processes requiring treatment at multiple resolutions, be it due to the need for a higher level of theory to capture relevant phenomena or because they unfold on considerable time scales. However, writing interfaces for such simulations involving several independent programs is always a balancing act between performance and flexibility.

Addressing this challenge, the MiMiC framework [1,2] provides a general interface for different programs handling individual subsystems at varying levels of theory. It facilitates data exchange between these concurrently running programs and potentially calculates interactions that might arise between the subsystems. To minimize the implementation and computational overheads, we adopted a loose-coupling paradigm combined with a multiple-program multiple-data model (MPMD). The key advantage of this approach is that MiMiC does not interfere with the underlying parallelization of different external programs, thus maintaining their high efficiency that can be fully exploited on modern modular high-performance computing architectures.

In recent years, MiMiC has undergone a major refactoring that further enhanced its flexibility, made it more accessible, and introduced several new programs (OpenMM, CP2K) into its ecosystem [3,4]. In this webinar, I would like to introduce the framework itself, with a particular focus on the software design, user experience, and performance of its generalized electrostatic embedding QM/MM scheme. Finally, I will provide a short overview of ongoing developments.

Speaker:
Andrej Antalík graduated in Chemical Physics at the Charles University in Prague, where he also obtained a PhD under the supervision of Prof. Jiří Pittner. In 2021, he joined the group of Prof. Ursula Röthlisberger as a postdoctoral researcher. His research focuses on the development of electronic structure and multiscale methods, and he is one of the main developers of MiMiC.