Skip to content

AlphaFold

Google DeepMind / Google DeepMind

AlphaFold is DeepMind's protein structure prediction system. The official repositories provide AlphaFold 2 inference code (Apache-2.0) and the AlphaFold 3 inference pipeline, with model parameters offered under CC BY 4.0 and separate non-commercial terms respectively.

Catalog updated ·

Overview

AlphaFold is the protein structure prediction system developed by Google DeepMind. This entry covers the official implementation repositories: the AlphaFold 2 repository (google-deepmind/alphafold) contains the inference code under the Apache 2.0 license, and a companion repository (google-deepmind/alphafold3) contains the AlphaFold 3 inference pipeline, also with Apache-2.0 licensed code.

AlphaFold 2 parameters are distributed under CC BY 4.0. AlphaFold 3 parameters are obtained directly from Google under the model parameter terms dated 2024-11-09. Those terms permit only non-commercial use by or on behalf of non-commercial organizations, restrict parameter redistribution, and prohibit using the parameters or output to train similar biomolecular structure prediction models. Output use is also subject to the linked output terms. AlphaFold Server and commercial cloud offerings have separate service conditions. The AlphaFold Database is a separate hosted collection of predicted structures. Terms: https://github.com/google-deepmind/alphafold3/blob/main/WEIGHTS_TERMS_OF_USE.md

The catalog separately lists OpenFold (an independent reimplementation) and structure-prediction systems such as Boltz and Chai-1; those remain distinct identities. Structure predictions are computational outputs, not experimentally determined structures.

Key Features

  • Official AlphaFold 2 inference code with documented installation (Docker, NVIDIA GPU, genetic databases).
  • Companion google-deepmind/alphafold3 repository implements the AlphaFold 3 inference pipeline.
  • Model parameters distributed separately: AF2 under CC BY 4.0, AF3 under non-commercial terms of use.

Use Cases

  • Predict protein structures from sequences as an input step for docking, modeling and analysis workflows.
  • Study the reference implementation used by reimplementations and downstream structure-prediction tools.

How to Use

Full AlphaFold 2 installation is documented in the repository and requires Linux, Docker with NVIDIA Container Toolkit, up to 3 TB of genetic databases and a modern NVIDIA GPU; model parameters are downloaded during installation. AlphaFold 3 parameters must be obtained directly from Google under its terms of use. See: https://github.com/google-deepmind/alphafold

Sources and verification

This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.

Repository source reviewed for this catalog version:

Read the editorial and verification policy

Related resources

alchemiscale

Open Source

alchemiscale is a high-throughput execution system for alchemical free energy calculations that distributes work across HPC, cloud, bare-metal and volunteer compute resources.

Open sourcePython

Computational Chemistry · Drug Discovery · Scientific Data

alchemlyb

Open Source

alchemlyb extracts alchemical free energy data from molecular dynamics output and estimates free energies with MBAR, BAR and thermodynamic integration on the PyData stack.

Open sourcePython

Computational Chemistry · Drug Discovery · Scientific Data

BioSimSpace

Open Source

BioSimSpace is an interoperable Python framework for biomolecular simulation that exposes multiple simulation engines behind one interface and can start, stop and monitor processes interactively.

Open sourcePython

Computational Chemistry · Drug Discovery

Chai-1

Model

Chai-1 predicts biomolecular complex structures involving proteins, small molecules and nucleic acids through local command-line and Python inference.

Open sourcePython

Computational Chemistry · Drug Discovery

cinnabar

Open Source

cinnabar analyses and plots free energy calculation results with consistent, reproducible statistics so that computational methods can be compared fairly.

Open sourcePython

Computational Chemistry · Drug Discovery · Scientific Data

ColabFold

Open Source

ColabFold makes AlphaFold2/AlphaFold3 protein folding predictions accessible via Google Colab notebooks and an installable command-line tool, with a built-in MMseqs2-based MSA service.

Open sourcePython

Computational Chemistry · Drug Discovery

Related guides

Agents

AI Agents for Chemistry: From Chatbots to Autonomous Research

A practical guide to eight chemistry and biomedical research agents: how tools, memory, planning and multi-agent roles work, which projects fit different tasks, and how to evaluate bounded autonomy with scientific oversight.