⚡Run Boltz-2 in a Workspace
Boltz-2 is an open-source tool used to model and predict biomolecular interactions. This tutorial will use example data to show you how to install and run Boltz-2 to generate complex structural models and binding affinities.
To run Boltz-2, please use a Workspace with at least 1 GPU to ensure you have sufficient processing power.
1. Install Boltz-2 & Dependencies
Navigate to the Workspaces tab and select an appropriate Workspace to run Boltz-2.
Your Workspace's hardware requirements must include at least 1 GPU. We also recommend using a Workspace with the Conda base image, which will already have Python 3 installed.
If you need help starting, updating, or creating an appropriate Workspace, please go to the Get Started with a Workspace tutorial to learn how!
Once the Workspace is running, find it in the Workspace list and click Open to launch your development environment.
Once your Workspace is open, open a terminal window in your development environment. You can do this either with the keyboard shortcut Control ` (control + backtick) or through the hamburger menu (three horizontal lines icon).
Once the terminal is open, we'll update packages and install Boltz-2 using the following commands. You should be able to copy and paste the following commands directly into the terminal and hit enter to run them.
sudo apt update
sudo apt install python3-pip
pip install boltz[cuda] -UYou may see an interactive command in the terminal asking whether you wish to continue updating dependencies, please answer Y/yes to continue. Installing Boltz-2 may take a few minutes to complete.
While Boltz-2 is downloading, click on the extensions icon (looks like a Tetris block) to open the extensions menu.
From the extensions menu, search for and install “Protein Viewer” by Arian Jamasb.
After we run Boltz-2, this will allow you to view the protein structure visualizations in the development environment.
2. Add and Prepare Example Files
Next we need the Boltz-2 sample file to run a test. Open the affinity.yaml file from the Boltz-2 GitHub repository, then select and copy the entire contents to the clipboard.
Once you have this copied, return to your development environment and create a new file in the base directory by clicking the New File button and typing the name affinity.yaml. This should automatically save and open your new file.
Now paste the saved text from the original file that you copied earlier into your new affinity.yaml file.
To run this test, we don't want to use the the MSA server for authentication, so we will add the text msa: empty on line 6. The file should look about like this (note that the sequence string is cut off in this screenshot).
Make sure that the msa: empty line is indented correctly! It should be on the same level as the protein id and sequence.
Run Boltz-2 & View Results
Copy and paste this command into your terminal window to run Boltz-2 on the sample file we just added.
boltz predict affinity.yaml --out_dir ./results --num_workers 0Boltz-2 may take a while to run as it is a large model, please be patient until it finishes!
If you run Boltz-2 in a Workspace with no GPUs, Boltz-2 will fail to complete and you will see the error message No supported gpu backend found!
If this happens, you need to stop the Workspace, save a Workspace Snapshot, and update the Workspace hardware requirements before restarting it. For help, please see Get Started with a Workspace: Configure Hardware Requirements.
When Boltz-2 has finished running, you should have a new directory in your Workspace called results/boltz_results_affinity with a nested directory called predictions/affinity that will contain your results.
To view a protein model in 3D, right click on one of the files like affinity_model_0.cif and select launch protein viewer from file(s)
And you're done! 🎉
Last updated