AlphaFold 3

AI model · Science / biology model by Isomorphic Labs

Predicting the structure and interactions of all of life’s molecules.

A scientific AI model developed by Isomorphic Labs and Google DeepMind that predicts the 3D structures and interactions of proteins, DNA, RNA, and ligands to accelerate drug discovery.

Model specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Launched
2024-05-08
Languages
English

Category: Drug Discovery

Domains
Drug discovery, molecular biology, pharmaceutical research
Released
2024-05-08
Target Molecules
Proteins, DNA, RNA, ligands, ions, modified residues
Primary Application
Rational drug design and biological research
Developer Collaboration
Isomorphic Labs and Google DeepMind
  • Predicts structures of proteins, DNA, RNA, and ligands
  • Jointly developed by Isomorphic Labs and Google DeepMind
  • Doubled accuracy on difficult novel biological systems
  • Enables virtual modification and binding prediction of molecules

Overview

AlphaFold 3 is a scientific AI model developed jointly by Isomorphic Labs and Google DeepMind to predict the structure and interactions of all of life's molecules. It extends beyond protein structure prediction to model DNA, RNA, ligands, and other biomolecules, providing a holistic view of biological complexes. For pharmaceutical and biotech researchers, it solves the problem of understanding how potential drug molecules bind to target proteins, especially for novel targets with limited existing data. The model allows scientists to virtually iterate on chemical designs and predict binding behavior before moving to laboratory synthesis. It serves as a foundational tool in the Isomorphic Labs Drug Design Engine to accelerate the discovery of new therapeutics. AlphaFold 3 is engineered to be accessible to scientists across disciplines, including those without specialized computational training.

Biological Modeling

  • Joint Structure Prediction: The model predicts the 3D structures and interactions of proteins, DNA, RNA, ligands, and other biomolecules within a single system.
  • Ligand-Protein Interactions: It models intricate networks of interactions between proteins and small molecules to understand how they inhibit or potentiate protein function.
  • Novel Target Discovery: The model predicts interactions for novel protein targets where very little is known in the literature or existing structural data.
  • Virtual Chemical Design: Researchers can virtually modify chemical structures and run predictions to see how changes affect binding with a target protein.

Research and Usability

  • Scientific Usability: The model is engineered for use by scientists without prior training in computer science or computational chemistry.
  • Cross-Modality Support: It supports innovative therapeutic approaches based on modulating protein-protein and protein-nucleic acid interactions.
  • Hypothesis Testing at Scale: The model allows teams to formulate and test multiple binding hypotheses simultaneously at scale using high compute power.

Use cases

  • Medicinal chemists virtually modify chemical structures to predict how they will interact with target proteins.
  • Biologists predict the 3D structure of novel protein targets where little literature or structural data is available.
  • Researchers explore new therapeutic approaches by modeling protein-protein and protein-nucleic acid interactions.

Specifications

Modalities
Biological sequences and ligand descriptions → 3D molecular structures and interaction maps
Domains
Drug discovery, molecular biology, pharmaceutical research
Published accuracy
More than doubled accuracy on the most difficult novel systems compared to previous versions
Released
2024-05-08
Target Molecules
Proteins, DNA, RNA, ligands, ions, modified residues
Primary Application
Rational drug design and biological research
Developer Collaboration
Isomorphic Labs and Google DeepMind
Performance Improvement
2x accuracy on difficult novel systems over previous benchmarks

How it works

  1. Input Biological Data: Researchers input biological sequences (proteins, DNA, RNA) and descriptions of small molecules or ligands into the system.
  2. Predict Structure and Interactions: The model generates a 3D prediction of the molecular complex, showing how the different components interact and bind.
  3. Iterative Design and Refinement: Scientists use the predictions to virtually modify chemical structures and refine their designs before moving to physical lab testing.

FAQ

Who developed AlphaFold 3?

AlphaFold 3 was developed jointly by Isomorphic Labs and Google DeepMind.

How is AlphaFold 3 used in drug discovery?

It is used to understand how potential drug molecules bind to target proteins, allowing scientists to iterate on designs virtually before laboratory synthesis.

Can scientists without a computer science background use it?

Yes, it is designed to be usable by scientists without prior training in computer science or computational chemistry.

Added by AntSpace from public sources (isomorphiclabs.com).

More AI models by Isomorphic Labs