Place: Molecular Dynamics Group, University of Groningen
Promotor: Prof. S.J. Marrink
Duration: 4 years
Start: july-dec 2009
Applications: email CV to s.j.marrink@rug.nl
Requirements: a solid background in biophysics; expertise in programming is useful.

Aim: To develop multiscale force fields and methodology that extends the accessible time and length scale for simulations of biomolecular systems while retaining atomic detail.

Background: The complexity of biomolecular processes warrants a description on more than just a single level (cf. Figure 1). The specific methodological challenge is to design effective simulation protocols for combining models describing multiple scales, i.e. multiscaling. In this PhD project we focus on combining the traditional fine grained, all-atom models of biomolecules to the MARTINI coarse grained (CG) force field, developed recently in our group [Marrink04, Marrink07, Monticelli08], in which small groups of atoms are united into effective particles. With such an approach, long time scale behavior can be explored at the CG level and, simultaneously, the FG model probes the most relevant states in more detail. The multiscale approach raises fundamental questions concerning how the different scales actually connect. In principle, two different multiscale methods can be distinguished, namely exchange and boundary methods. In the exchange method, particles actually change identity during the simulation, i.e. are converted from FG particles to CG particles and vice versa. Exchange multiscaling has recently been pioneered by the groups of Kremer [Praprotnik07] and van Gunsteren [Christen06]. Using a coupling parameter, model liquid systems were converted between FG and CG representations, either at spatial [Praprotnik07] or temporal resolution [Christen06]. A promising way to gain further sampling speed is to use the replica exchange method [Sugita99] to allow swapping between the coupling parameter dependent Hamiltonians while maintaining the correct ensemble statistics. Alternatively, a static approach can be taken in which a predefined part of the system is simulated in full atomic detail, and part of it at a CG level, analogous to QM/MM methods combining the AA with the quantum level. Exploratory work on boundary multiscaling methods has also been performed by several groups [Shi06, Neri05].

Tentative Workplan: In the first part of this project we want to develop a general protocol to convert configurations obtained at the CG level to the FG level, i.e. temporal resolution exchange. Several groups have shown how to do this for simple systems [Christen06, Praprotnik06], but we face a number of additional challenges for the specific use of our MARTINI force field. Problems for which solutions have to be found include the treatment of the solvent exchange (multiple FG solvent molecules are presented by a single CG site), and the efficient threading of a FG configuration into a CG site. A possible solution is to use distance restraints between the FG solvent molecules to form small clusters. To counteract the increase in density of the constrained solvent as a result of the loss of entropy, we will add an additional short range repulsion. For a robust threading algorithm we anticipate the use of a random insertion protocol for the FG degrees of freedom (DOF) which will be minimized with the CG DOF acting as constraints. Using a coupling parameter approach [Christen06, Praprotnik06] the CG interactions will be slowly turned off. Ideally this procedure will take place in a reversible manner; how this can be achieved is not clear at the moment. Work on this part will be done in collaboration with the group of Prof. W. van Gunsteren (ETH Zurich). The method will be tested on biologically relevant systems such as lipid bilayers and proteins.

The second part of the project consists of a novel variation of the boundary method. The novelty lies in the treatment of the interaction of the central, all-atom part, with the surrounding particles represented at the CG level: we would like to explore the use of virtual CG sites that couple the FG and CG DOF. These virtual sites are constructed from the center-of-mass of the underlying all-atom representation. The hybrid particles constituting the central molecule thus interact intra-molecularly using FG forces, and inter-molecularly according to the CG forces, as depicted in figure 2. The advantages are threefold: i) no specific CG-FG interactions need to be parameterized as is done in other boundary methods [Shi06], making the method easily applicable to any system of interest ii) no position dependent resolution transformation is required, therefore a well defined Hamiltonian exists without violating Newtonʼs 3rd law, two conditions not met with other approaches [Praprotnik06], and iii) the method naturally combines the advantages of FG models (accurate description of the molecule of interest) and CG models (explicit treatment of the surrounding solvent at a speed 2-3 orders of magnitude larger compared to FG models). The method appears ideally suited to study protein-ligand binding and protein folding, for instance. Pure CG models of proteins either require knowledge based interactions to study folding pathways, or use restraints which bias the accessible protein conformations to a particular (native) state. With the proposed novel boundary method these limitations are overcome. In collaboration with Prof. D. van der Spoel (one of the main GROMACS developers) the multiscale methodology will be implemented in the future GROMACS (4.x) releases. Applications of the proposed multiscale implementations are foreseen for many of the projects ongoing in our MD group, including membrane poration by antimicrobial and novel synthetic peptides, protein-protein complex formation, and many more. In summary, the major aims that we have in terms of multiscale development:

  • Develop efficient exchange strategies between FG models and the CG MARTINI model.
  • Develop a new hybrid CG/FG model suited for protein folding studies.
  • Integrate multiscale methodologies into the GROMACS software suite [Spoel05].

Embedding: The research will take place within the Molecular Dynamics group, attached to the Groningen Biomolecular Sciences and Biotechnology Institute (GBB). The MD group, headed by S.J. Marrink, concentrates on dynamical simulation of biopolymers and lipid aggregates. The aim is to understand and predict macroscopic behaviour of complex biomolecular systems on the basis of the effective interactions between atoms. The MD group is home to the new MARTINI coarse grained forcefield for biomolecular simulations. Furthermore, the MD group remains closely linked to the ongoing development of the GROMACS software, the origins of which are found in this group. The group currently consists of 12 students/postdocs from all over the world.

The PhD position is funded through a TOP-grant from the Netherlands Organization for Scientific Research (NWO). The PhD student employed on this project will collaborate strongly with the other members of our group, and internationally with the groups of W.F. van Gunsteren and D. van der Spoel.

For details see the document below: