gmx-template

Template for setting up MD simulations in the Harms lab.

Package contents

• example.py: example python file showing different ways to set up md runs.
• ff/: directory with forcefield parameters.
  • Models of various LPS and LPS-like molecules. Generated using the charmm-gui LPS modeler tool. These have names like e-coli_l-i-o-o1, which mean the LPS is taken from E. coli and has a lipid A (first l), inner core (i), outer core (o), and one O-antigen (o1).
  • charmm36-feb2021.ff. The CHARMM36m forcefield, ported to GROMACS. http://mackerell.umaryland.edu/charmm_ff.shtml#gromacs
  • mdp-files. Files used for setting up a GROMACS run.
• example-inputs/: directories with inputs used to show different ways to run example.py.
• example-outputs/: directory where outputs are written from example runs of example.py.

Build initial model

Prepping your protein pdb file

• Delete all HETATM except metal ions at key sites (e.g. CA in an S100). This includes deleting solvent.
• Renumber each chain so residues are different numbers in each chain.
  Not all MD analysis code knows about chains, meaning if you have chain A and B, each with residue 1-100, it can get confusing. I would make chain A go from 1-100 and chain B go from 1001-1100. You can do this in pymol (https://pymolwiki.org/index.php?title=Alter&redirect=no)
• Make sure TER entries break each chain.
• If you are docking a ligand, you might have to move sidechains using the pymol "Mutagenesis" wizard.
• If you are combining several structures into one, pose them relative to one another. I like the "Actions->Drag coordinates" feature in pymol.

Prepping LPS (or other custom ligands)

1. Generate MD parameters for LPS of interest by going to charmm-gui.org and using the "LPS Modeler Tool."
2. Download the resulting forcefield parameters.
3. Convert the forcefield parameters to gromacs format by going to charmm-gui.org and using the "Force field converter" tool. Upload the .psf, .crd, and all .str files.
4. Download the zipped file and uncompress it. Find the gromacs directory (which will have a bunch of files like .gro, .itp, and a toppar folder). Note where this ends up on your computer.
5. Obtain the matched forcefield in gromacs format (such as charmm36-feb2021.ff, which is included in this repo)
6. In pymol, open the .gro file from the gromacs directory pose it where you want in the protein structure.
7. Save out the posed LPS molecule as a .pdb file. MAKE SURE TO CHECK "Original atom order" and "Retain atom ids" when writing out.
8. If you are docking two or more LPS molecules, save out each pose as its own .pdb file.

Configure MD simulation

Setting up a run

1. Use scp to copy the following to talapas. Make sure these end up in /projects/harmslab/USERNAME.
   • The cleaned protein pdb file.
   • Any posed LPS pdb files.
   • The appropriate LPS gromacs directory
   • The gmx-template directory
2. On talapas, modify the copy and modify the gmx-template/example.py script appropriately for your system. This should be run within the gmx-template/ directory.
3. Start an interactive session on talapas (qsub -I -A harmslab).
4. Load gromacs:

module load intel/19
module load gromacs/2019.4

5. Run the copied_example.py script:

cd gmx-template
python copied_example.py

6. This script will generate an output directory in gmx-template. Let's say you said output_dir="my_output_dir" in the copied_example.py script. This will make a new directory called my_output_dir with several sub directories. Copy my_output_dir/final to some convenient location. I usually make a directory for each system I plan to simulate (MD-2-with-LPS, MD-2-apo, etc.) and then have replicate directories within each one (MD-2-with-LPS/run000, MD2-with-LPS/run001). The final directory you are copying would be run000, as this will be where the MD simulation actually runs. If you plan to do several replicate runs, you can copy final to run000, run001, etc. An independent simulation can then be run in each directory.

Starting a run

1. Go into your run directory (the copy of the final directory generated by copied_example.py, e.g. run000) and open the run_md.srun file using a text editor . Make sure the slurm bits at the top are correct--especially these bits.

#SBATCH --partition=longfat
#SBATCH --time=07-00:00:00
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=28

If you're not sure how to decide what these should be, talk to Mike.

2. In the run directory, open the md.mdp file. Most of this file should not be changed, however, check the following line:

nsteps                  = 50000000  ; 2 * 50000000 = 1000 ps (100 ns)

This sets how long the simulation will run. Assuming a 2 fs step (dt line below in mdp file):

$$run\ time = 2 \times 10^{-15} s \times nsteps$$

This means the default (50000000) will run for 100 ns:

$$100\ ns = 1 \times 10^{-7}\ s = 2 \times 10^{-15} s \times 50,000,000\ steps$$

If you don't know how long you need to do your run for, talk to Mike.

3. Start the run by typing:

qsub run_md.srun

4. Once the run starts, it will sequentially make the following log files: em.log, nvt.log, npt.log, and md_0_1.log. You can follow the progress of the run by typing tail -f BLAH.log. (You'll have to hit CTRL+C to exit the tail call). The vast majority of the run will be occupied by the md run part of the script.
5. The run will go until it either completes all steps or you run out of run-time, at which point slurm will kill your job.

After the run

Extracting your run into useful compressed file

On talapas, navigate to the directory you ran the simulation in and type qsub extract_md.srun md_0_1 npt.gro output. This will make a directory called output that has three files: traj.gro, traj.tpr, and traj.xtc. These contain the main outputs from your run for further analysis. Copy this whole directory down to spock or your personal computer for analyses. You can run extract_md.srun this while the run is in progress. It will simply extract up to however far the run has progressed.

Extending a run

If you want to restart a stopped job--whether because it ran out of cluster time or because it finished all specified steps--edit the run_md.srun file. There are a few lines commented at the bottom of the file that give instructions for how to do this.

Visualizing results

Unless your runs are short and/or your computer is powerful, you'll likely need look at the runs sitting at spock. You'll want to use the program vmd. Navigate to wherever you download the simulation and type vmd traj.gro traj.xtc. A vmd tutorial is here: http://www.ks.uiuc.edu/Training/Tutorials/vmd/tutorial-html/index.html