Preparing for MD simulation

This document shows how to prepare the cluster centers we get from FeatureSelection module to run an MD simulation. A few things to consider are as follows:

  1. Prepare the protein structure: The structures acquired from AlphaFold2 does not contain hydrogens. Hydrogens need to be added to the structures. Also, water and salt need to be added to the simulation box.

  2. Monitor the CVs: The selected a few hundred CVs need to be output throughout the simulation. We need to know their atom indices. Note that these indices will change after hydrogens are added to the protein structure.

Prepare the protein structure

The af2rave.simulation.utils.SimulationBox takes care of solvation and generating a MD-ready structure. A sample code is

import af2rave.simulation
box = af2rave.simulation.SimulationBox("sample.pdb")
box.create_box(ionicStrength=0.15)
box.save_pdb("sample_box.pdb")

You can provide other parameters to the create_box method, such as how thick the water padding should be, the type of water model (default: TIP3P) and salt (default: NaCl).

Caution

By default the package uses CHARMM36m force field. If you want to specify a different one, you need to pass a OpenMM ForceField class to the SimulationBox constructor. See the documentation for details.

Monitor the CVs

The pairwise distance CVs are represented by a list of tuples, where each tuple contains two atom indices. The monitored CVs can be read from the FeatureSelection module. For example:

indices = [tuple(fs.atom_indices[n]) for n in names]

However, after adding hydrogens, these indices will change. The SimulationBox object created can take care of this change:

import numpy as np

new_indices = box.map_atom_index(indices)
np.save("cv_indices.npy", new_indices)

This saved file can be subsequently transferred to your computer cluster for MD simulation.