This repo contains a Jupyter notebook to generate random landmarks of a protein strucure, derive appropriate collective variables, and run metadynamics simulation of the protein unfolding-folding.
The notebooks must be run in a Docker container with all the Python dependencies and a prebuilt Gromacs + Plumed separate image downloaded from Dockerhub. Dockerfile defining such container is provided.
Build the container yourself with the provided build.sh script and run it with run.sh (eventually modifying the container name and port to listen).
If nvidia-docker is available, it is used and GPU acceleration works. The current version works with Cuda 10.x only.
The working directory is mounted into the container, as well as /var/run/docker.sock, so that a sibling container (running gromacs) can be invoked from inside. The notebook creates a subdirectory derived from the PDB input file name, and puts all results there.
In either case, go to index.ipynb and enjoy!
A. Křenek, J. Hozzová, J. Olha, D. Trapl, V. Spiwok, Exploring Protein Folding Space with Neural Network Guided Simulations, Proc. ESM 2020, EUROSIS-ETI Publication, pp. 305-309 ISBN: 978-9492859-12-9.