Skip to content

README.md

Latest commit

 

History

History
122 lines (83 loc) · 3.11 KB

README.md

File metadata and controls

122 lines (83 loc) · 3.11 KB

Docker Compose

In this directory you will find a Docker Compose yaml file that will start up the NIMs for AlphaFold-2, MolMIM, and DiffDock. Once the NIMs are started, you will probably then want to take a look the example notebook in the src directory.

Prerequisites

You will need to install:

Get Your API Key

Navigate to the API Catalog, ensure you're logged in, and click "Download Blueprint" to generate an API Key.

It can be convenient to save your API Key to a file for later use. Here, we'll assume you've saved your API Key to a plain text file called ~/API_KEY.

Set an environment variable named NGC_CLI_API_KEY to your API Key.

export NGC_CLI_API_KEY=$(cat ~/API_KEY)

Login to nvcr.io

To pull the containers used in the docker compose configuration, you will need to login to nvcr.io with your NGC API Key.

$ docker login nvcr.io
Username: $oauthtoken
Password: <PASTE_API_KEY_HERE>

Create Cache Directories

You'll need to create at least one directory for model files, weights, and MSA databases. The simplest way to go is to do

mkdir -p ~/.cache/nim
chmod -R 777 ~/.cache/nim

Optional: Specify An Alternate Directory

The protein sequence databases necessary for the MSA step in AlphaFold-2 require about 500 GB of disk space. It can be useful to specify another file location for AlphaFold-2 that will get used to store these files. For example, if you had a larger filesystem at /scratch, you might do:

mkdir -p /scratch/nim
chmod -R 777 /scratch/nim

Then you would want to set the environment variables $ALPHAFOLD2_CACHE, $DIFFDOCK_CACHE, and $MOLMIM_CACHE as follows using the example directory we just created.

export ALPHAFOLD2_CACHE='/scratch/nim'
export DIFFDOCK_CACHE='/scratch/nim'
export MOLMIM_CACHE='/scratch/nim'

Start Up the NIMs

Starting the NIMs is as simple as running

docker compose up

You should see Docker pulling the containers, which can take up to several hours, depending on the speed of you internet connection. Once the containers are pulled, they will start. It will take ~5 minutes to start the three NIMs.

Check the Status of the NIMs

Check AlphaFold2 status:

curl localhost:8081/v1/health/ready

Example output:

{"status":"ready"}

Check DiffDock status:

curl localhost:8082/v1/health/ready

Example output:

true

Check MolMIM status:

curl localhost:8083/v1/health/ready

Example output:

{"status":"ready"}

Interacting with the NIMs

The three NIMs are available on three different ports:

  • AlphaFold2: 8081
  • MolMIM: 8082
  • DiffDock: 8083

For a complete example, navigate to the src directory where you'll find a Jupyter notebook with a complete generative virtual screening pipeline.