jourmore/openmm.py

Background

https://github.com/pablo-arantes has developed an excellent Colab workflow to run the simulation. However, Google Colab sometimes has connection issues or you have to subscribe to get access to a decent GPU, or in some areas, it is just not accessible to connect to Colab easily.

This repo is designed for people who want to do a general molecular dynamic simulation locally using OpenMM with Amber inputs, with much more stability.

Pre-requirement

OpenMM need to be installed on a Linux platform

Steps

generate receptor.pdb and Docked1.pdb, or with Schrodinger/Maestro, or Autodock vina, or any other docking software you could have access to. Alternatively, you can always to install webdock (https://github.com/quantaosun/webdock) into your local Linux, and then generate the docked results. Updated: A better way to do the docking is https://github.com/quantaosun/labodock_binder, the stable version is the preferred link.
, click to wait loading finished. Alternatively, you can install this repo, which comes with amber tools (https://github.com/quantaosun/pl3_gmx_mmpbsa) into your local Linux, then generate the md inputs files as described in step 3. If you prefer something more stable and more reliable, you could use the docker image on local Windows to do so.
Upload the 2 files from step1, to generate SYS_gaff2.prmtop and SYS_gaff2.crd, note, the binder link is not guaranteed if there is a high visit volume, if not accessible you can use https://github.com/pablo-arantes/making-it-rain
Download this repo to a local Linux with OpenMM installed (If you haven't got one, you could create one by the environment file attached in this repo), and copy the two files in step2, to the same folder as the Python script provided in this repo.

Inputs

prot_lig_openmm.py
SYS_gaff2.crd
SYS_gaff2.prmtop

Execute the simulation

python prot_lig_openmm.py It will do a 10 ns equilibration and 50 ns of production by default. You must change the prot_lig_openmm.py if you want a longer or shorter simulation.

Output

prod_lig_1.pdb
prot_lig_prod.dcd
prot_lig_prod.log
stdout

prod_lig_1.pdb is the final simulated structure, prot_lig_prod.dcd is the trajectory file.

Visualisation

In Pymol, load prod_lig_1.pdb and prot_lig_prod.dcd

Performance

RTX4090: to simulate 100 nanoseconds for a typical kinase size protein-ligand complex

#"Progress (%)"	"Step"	"Time (ps)"	"Potential Energy (kJ/mole)"	"Temperature (K)"	"Speed (ns/day)"	"Time Remaining"
0.0%	1000	8002.000000712723	-768173.560093076	300.0716998151275	0	--
0.0%	2000	8004.00000071313	-767249.5673865234	300.76147916830683	486	    59:13
0.0%	3000	8006.000000713538	-767389.9470499307	301.14007459911784	486	    59:13
0.0%	4000	8008.000000713945	-768387.7632756401	298.08645047906015	486	    59:14

Simulation time

I provided three potential lengths of simulation, you can open prot_lig_openmm.py to comment or uncomment lines to choose one of the

long, 20 ns equilibration + 100 ns production
medium, 10 ns equilibration + 50 ns production
short, 2 ns equilibration + 10 ns production

simulations. Like in the section below, I chose the short one by uncommenting those lines.

# Simulation Options

# long simulation (default)
#minimizationSteps = 10000
#productionSteps = 50000000  # 100 ns
#equilibrationSteps = 20000000  # 20 ns
###################################################
# medium simulation
#minimizationSteps = 10000
#productionSteps = 25000000  # 50 ns
#equilibrationSteps = 10000000  # 10 ns
###################################################
# short simulation
minimizationSteps = 10000
productionSteps = 10000000  # 10 ns
equilibrationSteps = 4000000  # 2 ns
##################################################

Analysis

For overall MD quality please see the post-analysis folder for instructions.
For protein-ligand interaction analysis, You could use the docker image mentioned in the repo, and analyse the results. Or if you just want to analyse the final PDB complex, you could just use Maestro, Free Maestro etc to generate the 2D interaction diagram.

jourmore / openmm.py