Skip to content

AlphaFold 3

Resource Location

/labs/SysMedBio/Q/Resources/AI/AlphaFold3

/Volumes/lab/SysMedBio/Resources/AI/AlphaFold3

Q:\SysMedBio\Resources\AI\AlphaFold3

AlphaFold 3 is a revolutionary structural biology tool that predicts the 3D structure and interactions of proteins, nucleic acids (DNA and RNA), chemical modifications, and small molecule ligands.

Scientific Context

AlphaFold 3 represents a major paradigm shift in biomolecular modeling. While its predecessor, AlphaFold 2, focused primarily on single protein structures or protein-protein multimer complexes, AlphaFold 3 extends this capability to model complete biomolecular complexes across a wide spectrum of biological entities.

Key Capabilities

  • Complex Modeling: Predicts coordinates for assemblies containing proteins, DNA, RNA, small molecule ligands, ions, and chemical modifications (such as phosphorylation, glycosylation, or methylation).
  • Biological Scope: Highly useful for drug discovery (protein-ligand interactions), transcriptional regulation (protein-nucleic acid interactions), immunology (antibody-antigen complexes), and epigenetics.
  • Architectural Innovations:
    • Tokenization: Represents standard amino acids and nucleotides as sequence tokens, while modeling ligands and ions at the atomic level.
    • Pairformer Module: Replaces Evoformer with a simpler, faster network to process representation pairs, reducing computational overhead.
    • Diffusion Structure Module: Replaces the traditional structure module with a generative diffusion model that refines coordinates from random noise, eliminating the need for post-prediction physical refinement (e.g., AMBER minimization).

Main Publication

  • Title: Accurate structure prediction of biomolecular interactions with AlphaFold 3
  • Authors: Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvunakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, Charles Beattie, Ottavia Bertolli, Alex Bridgland, Alexey Cherepanov, Miles Congreve, Alexander I. Cowen-Rivers, Andrew Cowie, Michael Figurnov, Fabian B. Fuchs, Hannah Gladman, Rishub Jain, Yousuf A. Khan, Caroline M. R. Low, Kuba Perlin, Anna Potapenko, Pascal Savy, Sukhdeep Singh, Adrian Stecula, Ashok Thillaisundaram, Catherine Tong, Sergei Yakneen, Ellen D. Zhong, Michal Zielinski, Augustin Žídek, Victor Bapst, Pushmeet Kohli, Max Jaderberg, Demis Hassabis & John M. Jumper
  • Journal: Nature (2024, Vol. 630, Issue 8016, Pages 493–500)
  • DOI: 10.1038/s41586-024-07487-w

Available Files

The following key files and folders are available in the AlphaFold 3 resource directory:

  • s41586-024-07487-w.pdf: The official publication PDF.
  • af3.bin.zst: Compressed model parameters and database archive.
  • alphafold3/: Repository codebase containing:
    • run_alphafold.py: Main entry point script for running prediction pipelines.
    • run_alphafold_test.py & run_alphafold_data_test.py: Pipeline verification tests.
    • fetch_databases.sh: Utility script to fetch required databases.
    • pyproject.toml & uv.lock: Python environment package configuration and lockfiles.
    • docker/: Docker container configuration and execution scripts.
    • docs/: User guides, installation instructions, input/output specifications, and benchmark metrics.
    • src/: Core model source code.
    • WEIGHTS_TERMS_OF_USE.md & WEIGHTS_PROHIBITED_USE_POLICY.md: Legal usage constraints for model parameters.
    • OUTPUT_TERMS_OF_USE.md: Legal terms for usage of predictions.