table of contents
Introduction
AlphaFold3 has a highly accurate structure prediction function and has attracted a great deal of attention in the field of protein structure prediction.
However"I want to actually try it outWhen thinking about this, many people feel that there are hurdles to implementing things like GPUs and containers.
In this article, we will introduce how to install it on Ubuntu 24.04.Organizing the overall picture in three stepsWe will also introduce the efforts of TEGSYS, which supports the implementation.
AlphaFold3 installation guide now available as a technical article
TEGSYSSo, in the Ubuntu 24.04 environmentAlphaFold3 installation procedureThis technical article provides specific example commands and configuration steps to help you set it up.
We also explain points to be aware of when implementing the system and areas where people may get stuck.
Overview of implementation (3 steps)
The process of launching AlphaFold3 from scratch can be summarized in three steps:
- Step 1: Prepare your environment
We will introduce Docker, CUDA Toolkit, and NVIDIA Container Toolkit.
nvidia-smiCheck that the GPU is correctly recognized by running the command: - Step 2: Install AlphaFold3
Get the official repository and build the Docker image.
Models can be obtained by applying through Google Forms.$HOME/modelsPlace it in.
We also acquire and deploy the necessary databases. - Step 3: Run and check
Prepare an input JSON file and run it with a specified GPU.
If there are no errors in the log, the deployment is complete.
If you keep these three steps in mind, it will be easier to grasp the overall picture of the implementation.
For example commands and important points for each step, please see the TEGSYS article.
For those who want to roughly understand the guidelines for the installation environment
Once you understand the installation procedure for AlphaFold3, the next thing you should check is "How much preparation is needed to actually get it working??” is the question.
Here we have clearly organized the elements required to create an environment in which AlphaFold3 can be run.
Understanding the overall picture will lead to stable operation. Please make use of it.
| Category | Required elements |
|---|---|
| OS | Ubuntu 24.04 |
| Container environment | Docker, NVIDIA Container Toolkit |
| GPU environment | CUDA (driver matching required) |
| Database | Compressed 252GB / Expanded approx. 630GB (with acquisition script) |
| モデル | Obtained after applying for Google Form,$HOME/models Placed in |
This configuration is based on an Ubuntu 24.04 environment verified by TEGSYS, allowing for highly reproducible setup even for first-time deployments.
By basing it on a verified environment, problems during implementation can be prevented and operation can be ensured smoothly.
TEGSYS supports implementation and operation
Once all the necessary elements for building the environment are in place, it will be possible to run AlphaFold3 inference, but in actual operation, you may encounter hardware bottlenecks.
there TEGSYS provides implementation support that takes into account these unique challenges in research settings.
We provide consistent support from environment construction to stable operation, including GPU performance and dependency matching.
Next, we will introduce an example configuration that assumes AlphaFold3 operation.
Reference example of actual machine configuration
Below is an example of a workstation configuration for running AlphaFold3.
It can be flexibly customized according to the actual application and analysis target.
Applications: Alphafold3, ESM Cambrian, Machine Learning, Gaussian
| CPU | AMD Ryzen9 9950X 4.30GHz (Boost max 5.70GHz) 16C/32T |
| memory | Total 192GB DDR5-5600 (48GB x 4) |
| storage | 1TB SSD (SATA) + 4TB SSD (M.2 NVMe Gen4) |
| GPU | NVIDIA GeForce RTX 5090 32GB |
| OS | Ubuntu 24.04 |
* NVMe was adopted in consideration of large-capacity database deployment and I/O. The memory configuration also takes into account multiple users and parallel processing.
Applications: RNA-seq analysis, genome analysis, structure prediction
| CPU | Intel Xeon W7-2595X 2.80GHz (up to 4.8GHz at TB3.0) 26C/52T |
| memory | Total 256GB DDR5-5600 REG ECC (64GB x 4) |
| storage | 2TB SSD (SATA) x 2 (no RAID) + 4TB SSD (M.2 NVMe Gen4) |
| GPU | NVIDIA RTX 4500 Ada 24GB (DisplayPort x 4) |
| OS | Microsoft Windows 11 Pro for WS 64bit (Ubuntu 22.04 dual boot) |
* While taking into consideration Windows-based operations, a Linux environment is also secured through dual boot. Switching between these environments can be flexibly performed according to the analysis workflow.
It is designed specifically for research purposes and can be used immediately after installation. It can also be delivered pre-installed.
If you are unsure about the environment setup or settings,You can consult us.
Life Science Campaign Announcement
In addition to this implementation support, TEGSYS is currently running a special campaign for researchers in the life sciences field.
For those who are considering introducing a real machine configuration,Up to 10TB of internal HDD is provided free of charge as a storage serviceWe offer more practical benefits such as:
The campaign page also publishes examples of the implementation of various software and analysis environments.
Please take a look at this as a reference for environment construction and system design.
Summary
The implementation of AlphaFold3 may seem complicated at first glance, but if you understand the key points,Stable operation is possible.
By preparing the GPU environment and container settings,Highly reproducible structure prediction environmentRealizeAvailable
At TEGSYS,Supporting the development of cutting-edge research infrastructureWe are promoting the creation of a foundation that will take our customers' research to the next stage.
In addition, We are running a special campaign to encourage researchers in the life sciences field to adopt our technology.
Please see below for details.
