A collection of scripts to equilibrate and perform production runs of peptides using ANI and AMBER
Developed on Ubuntu 20.4 with Nvidia driver 510 and CUDA 11.6
Tested working on Ubuntu 20.4 Nvidia driver 495 and CUDA 11.5
Minimum CUDA 11.4 required
git clone https://github.com/meyresearch/ANI-Peptides && cd ./ANI-Peptides
conda env create -n ani_test -f environment.yml
git clone https://github.com/meyresearch/openmm-ml
pip install ./openmm-ml/.
Key dependencies:
- OpenMM
Base molecular simulation toolkit
- TorchANI
PyTorch Implementation of ANI
- PyTorch
TorchANI runs on the PyTorch machine learning framework
- OpenMM-Torch
A plugin for OpenMM that allows PyTorch static computation graphs (TorchANI) to be used in OpenMM as a TorchForce object, an OpenMM Force class
- OpenMM-ML
Implements TorchANI as an OpenMM TorchForce using OpenMM-Torch. The glue that brings everything together!
- MDAnalysis
Trajectory analysis library
- Seaborn, Matplotlib, Pandas
Data plotting and manipulation
In general, you will want to modify the scripts equilibration.py and production.py to perform simulations in a consistent way, then execute the scripts on your pdbs using the available command line arguments
Example: python equilibrate.py pdbs/aaa.pdb amber
This will perform equilibration of aaa.pdb using AMBER forcefield. For more options, modify equilibration.py script.
The results will be saved in /outputs/equilibration_aaa_amber_xxxxxx_xxxxxx, where xxxxxx_xxxxxx is a timestamp
- State data (csv)
- Graphs of state data (pdf)
- Save state of equilibrated system (xml)
- Save state of equilibrated system (pdb)
Example: python production.py pdbs_equilibrated/aaa_amber.pdb amber
This will perform a production run of the equilibrated aaa_amber.pdb using AMBER forcefield. For more options, modify production.py script.
The results will be saved in /outputs/production_aaa_amber_xxxxxx_xxxxxx, including
- State data (csv)
- Graphs of state data (pdf)
- Frames of the production run saved at regular intervals (dcd)