These scripts are a first attempt to do reverse mapping or fine-graining where a bead model from mesoscale is converted back into the atomistic representation. The methodology is based on a paper by Marrink et al. where they used a mixture of restraints and simulated annealing to gradually relax the geometry of the atomistic structure. For this, we have created two scripts, one to create the raw atomistic structure (FineGrainPeriodicV1.pl) and the second to relax the structure (RelaxFineGrainedAtomsV1.pl). This post details what the scripts do, a brief description of the example system included in the zip file, and some comments and caveats.

FineGrainPeriodicV1.pl

This script creates the initial atomistic structure from the bead structure. It requires that the mapping between the beads and atomistic structure is completed before the bead structure is generated (ie you really need to think about whether you are going to fine-grain before you run your mesoscale simulations!) This mapping is done by creating centroids and motion groups on the initial atomistic structure. The names of these centroids and beads in the atomistic model should map to the names of the beads in the coarse-grained model.

The script identifies each mesomolecule and copies the atomistic molecule into fine grained document. The atomistic molecule is mapped onto the same positions as the mesomolecule by setting the centroid positions on the atomistic model to the bead centres of the mesomolecule.

Also, an anchor point is created at the centre of mass of the bead and a restraint is created between the centroids on the atomistic molecule and the anchor point.

The output of this script is a raw atomistic model which then needs relaxation.

RelaxFineGrainedAtomsV1.pl

The second script is designed to relax the atomistic structure into a more realistic configuration. This script uses three stages:

• Optimization: An initial geometry optimization is performed with a high restraint energy between anchor point and centroids on the molecule. This is designed to smooth out the initial bad geometries.
• Annealing: A high temperature annealing run is performed starting at 1300K and dropping to 300K. The restraint is still maintained at the maximum value.
• Relaxation: An NVT dynamics is performed at 300K and the restraint energy is gradually relaxed in a series of steps.

Example Input Structures

The attached files show reverse-mapping for a DPPC lipid in water. In the MARTINI forcefield, a bead of water represents four water molecules. To create this, we optimized four water molecules in a cluster and placed a centroid indicating the centre of mass of the beads. As a very high temperature molecular dynamics is used in the relaxation process, we also included restraints from the oxygen on each water molecule to the centroid for the water bead. This should decrease artificial diffusion of the water molecules caused by the high temperature. See below for the definition of the water bead.

The DPPC lipid was coarse-grained and fine-grained using the standard approach defined in the first MARTINI paper by Marrink et al. To create this, we initially defined centroids on each group of atoms representing a bead and then created the motion groups. The name of the motion group and centroid for each “bead” of atoms needs to match and but each “bead” of atoms needs a unique name. For example, see the structure below where each motion group is labelled by its name.

Comments and Caveats

• These are experimental scripts and you should expect to be modifying the settings in the relaxation section to suit your system. You need a good understanding of mesoscale and classical simulations to use these scripts.

• You should design your mesoscale calculation with fine-graining in mind - it is not very easy to apply this to an existing set of simulation results unless you have though carefully about bead names (ie given each bead in a mesomolecule a unique name).

• The restraint energy chosen for this example may need to be modified. We have not spent any time playing around with these values.

• The relaxation script will take several hours to run and should be run in parallel.

• You are advised to read the paper by Marrink on their experiences using this approach.

Let us know how you get on with this script!

Regards,

Reinier and Stephen