Binding free energy calculations pipeline on GCP with Gromacs

This repository will contain a pipeline developed by Google Research, the Max Planck Institute for Multidisciplinary Sciences, and Janssen, to carry out protein-ligand binidng free energy calculations using Gromacs and Google Cloud Platform (GCP).

Disclaimer

This is not an officially supported Google product. The repo is also not under active development.

Content

• colab_tutorial contains Colab notebooks with simple examples on how to run Gromacs and binding free energy calculations in Colab, as well as how to use xmanager to submit jobs in GCP.
• docker contains files that can be used to build Docker images with Gromacs and pmx.
• xcloud contains example of how to run simulations, as well as more complex absolute binding free energy (ABFE)calculation protocols, on Google Cloud using Vertex AI.

Usage

The main script defining the ABFE protocol is xcloud/abfe_base_protocol/main.py. This is based on a non-equilibriunm protocol, similar to that proposed by Khalak et al..

A job can be submitted to GCP with the script xcloud/abfe_base_protocol/launch_vertex.py and using xmanager. To be able to submit the calculations on GCP, you need to have a GCP project and billing account set up. In addition, a docker image with both Gromacs and the pmx Python library should be available in the Artifact Registry.

Once all the above is ready, a calculation (or a batch of calculations) can be submitted as follows:

xmanager launch launch_vertex.py -- \
  --base_image="gcr.io/$MY_PROJECT/$PATH_TO_DOCKER_IMAGE" \
  --top_path="/$GCS_PATH_TO_TOP_FILES" \
  --mdp_path="/$GCS_PATH_TO_MDP_FILES" \
  --out_path="/$GCS_OUTPUT_PATH" \
  --lig_ids="1,10"

Where base_image refers to the Docker image, and top_path, mdp_path, and out_path refer to paths in Cloud buckets. For additional information, see the description of the script's arguments.