AlphaFold3
Last Updated: June 2, 2025
Module
| Module Name | Version |
|---|---|
| alphafold | 3.0.1 |
This module is not required for execution. The module help command can be used to display information about a sample job script.
Overview
AlphaFold3 is a software developed by DeepMind that uses machine learning techniques to predict the three-dimensional structure of proteins and other molecules.
- Development repository google-deepmind/alphafold: Open source code for AlphaFold 3.
How to use
Unlike AlphaFold2, AlphaFold3 clearly separates the model parameters from the sequence/structure database. The sequence and structure databases are placed on Genkai's high-speed storage for reference, but the model parameters must be obtained by the users themselves. To obtain them, you need to apply for them through Google Forms. A sample job script to run AlphaFold3 using the prepared database file is provided below. Copy the file to your home directory and modify it as necessary. Since it seems that using multiple GPUs does not seem to improve performance, the job is designed to use only one GPU.
- /home/center/app/alphafold/v3.0.1/samples/job_sample.sh
The scripts that are called at the end of the script and actually executed are as follows
- /home/center/app/alphafold/v3.0.1/run_alphafold.py