Molecular Simulation Community

Molecular Simulation Community

 

Introduction

 

Semi-empiric simulation methods have been responsible for tremendous advances in chemistry, life science, and materials research. The possibility to accurately model finite temperature, time development, and very large systems has made these methods a cornerstone for biomolecular modeling and drug design, it has become possible to simulate complex processes such as protein folding on millisecond scale involving >100,000 atoms, and there is a large international effort (OpenKIM) to standardize models of interatomic potential for materials research.

The common theme for these applications is that the main challenge lies in correctly modeling entropy of the system, and the simulations build on statistical mechanics concepts to efficiently sample phase space. In contrast, the interactions potentials are frequently relatively simple in order to sample billions of different conformations, e.g. protein structures or an inkjet droplet being absorbed by a polymer paper surface.

Over the last decades, considerably efforts have been spent on improving parameters and semi-empiric models to better reproduce experimental properties, and the modeling codes have been rewritten to run in parallel on thousands of cores to enable simulations to reach longer timescales. This has led to very important advances for actual applications: we can now typically predict the free energy of binding or solvation for a small molecule to within 1kJ/mol, and molecular simulation has become a standard additional method in many high-impact life science papers.

The field is in rapid development, and faced with a number of challenges. It is critical to develop new methods that can scale molecular simulation methods to future exascale supercomputers and efficiently use new streaming architectures such as GPUs. Simultaneously, there is important fundamental research on how to better sample complex systems, in particular the development of Markov State Models and simulated tempering algorithms that make it possible to use thousands of loosely coupled simulations, e.g. in distributed computing. It is also important to develop multiscale methods to combine e.g. quantum chemistry, atomic simulation, coarse-grained simulation, and continuum modeling as well as a better integration between bioinformatics and molecular modeling. These projects require close collaborations with the SeRC communities on Numerical Analysis, Electronic Structure and Bioinformatics.

 

 

Research environment

 

SeRC hosts a number of leading groups on molecular simulation. The division of Theoretical & Computational Biophysics (TCB) at KTH is led by Erik Lindahl (Professor, also affiliated with SU) with Berk Hess as additional tenured staff.  TCB is supported by the Swedish Foundation for Strategic Research, Lindahl was awarded an ERC grant in the first round, and Hess another one in 2010. TCB is part of the Center for Biomembrane Research as well as Stockholm Bioinformatics Center, and has close collaborations with groups at the Science for Life Laboratory, another strategic research area run by KTH/KI/SU. The division is the main site for development of the GROMACS simulation toolkit, which is also the engine of the Folding@Home project. Additional groups at KTH include Olle Edholm (Professor) at Theoretical Physics working on membrane simulations, Mats Wallin (Professor) working on statistical physics, Anatoly Belonoshko (Associate Professor) working on materials modeling, and Anders Szepessy (Professor) at Numerical Analysis working molecular dynamics algorithm development.

The Molecular Modeling group at KI is headed by Lennart Nilsson (Professor), who is one of the core developers of the CHARMM molecular simulation package, with close ties to the Karplus group at Harvard, and has an extensive research program focused both on method development and simulations of DNA and ligand binding.

The department of Materials and Environmental Chemistry at SU hosts Aatto Laaksonen (Professor), Alexander Lyubartsev (Professor), and Arnold Maliniak (Professor), working both on materials and multiscale modeling as well as biomolecular simulations.

In particular for biomolecular modeling the Stockholm-Linköping area provides a strong extended environment through several experimental collaborations with SU and Linköping Hospital, and there are also close contacts with groups working on sequence-based bioinformatics, with a number of joint publications. Several of the Stockholm groups also have close collaborations with researchers in Uppsala, in particular Johan Åqvist (Professor) and David van der Spoel (Professor), Lynn Kamerlin (Associate Professor), as well as Samuel Colburn-Flores (Associate Professor). 

 

 

Overall research goals

 

The unifying challenge for our community is to make particle-based simulation methods more predictive and quantitative, both for materials and biomolecular systems, based on statistical mechanics. SeRC makes it possible to approach this together with experts from other communities as well as HPC infrastructure, and our prioritized aims are to:

  • Enable molecular simulation methods to scale efficiently over tens to hundreds of thousands of cores, since this is critical to be able to use next-generation hardware and model large biomolecular systems.
  • Develop improved sampling techniques to make it possible to sample complex processes even on scales where it is simply not possible to study with a single simulation – for instance large protein folding.
  • Bridge scales in length and time, all the way from quantum chemistry to coarse-grained and continuum models.
  • Model effects of mutations and protein interactions by combining bioinformatics and molecular simulation in a high-through approach to genetic data at the Science for Life Laboratory in Stockholm.