Cancer3.AIAI in OncologyAI Models › AlphaFold 3
Foundation model Gated weights (licence click-through) Not applicable protein-structure

AlphaFold 3 3

Diffusion-based successor that predicts joint structures of proteins with DNA, RNA, ligands, ions and modified residues — the interaction types that matter for drug design and for understanding oncogenic complexes.

At a glance

DeveloperGoogle DeepMind / Isomorphic Labs
Version3
Released2024-05-08
LicenceCC-BY-NC-SA 4.0 (code); model parameters under academic non-commercial terms
AvailabilityGated weights (licence click-through)
KindFoundation model
Regulatory statusNot applicable

What it does

Models complexes (protein–ligand, protein–nucleic acid, antibody–antigen) in one network; AlphaFold Server offers free non-commercial predictions, and the code and weights (on request) allow local runs.

Tasks, data types and cancers

CancerPan-cancer
Inputsequences of all chains, ligand SMILES/CCD codes, ions
Outputall-atom coordinates of the complex, confidence metrics

Architecture

FamilyPairformer + diffusion structure module
Pre-trainingsupervised on PDB complexes + distillation

Training data

Protein Data Bank complexes including ligands and nucleic acids (cut-off 2021), with distillation data as in AlphaFold 2.

InstitutionsDeepMind, Isomorphic Labs

Linked datasets

Evaluation

Benchmark / datasetMetricValueExternal validationSource
PoseBusters (protein–ligand) % poses within 2 Å RMSD substantially above docking baselines (paper) yes Source

How to run

LibraryJAX (repo)
HardwareGPU with ≥ 80 GB recommended for large complexes

Model parameters are granted for academic use on request (form linked from the repo); commercial use requires a licence from Isomorphic Labs. The Server is the zero-setup route.

Regulatory status and intended use

Regulatory statusNot applicable
Intended useResearch tool.

Source →

Regulatory status is quoted from the source linked above and can change. Research-use-only models must not be used for clinical decisions.

Limitations and bias

  • Hallucinated structure in disordered regions; occasional chirality/clash errors flagged by the confidence outputs.
  • Non-commercial weights; commercial drug-discovery use is licensed separately.

Sources

  1. Abramson J et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 2024
  2. GitHub — google-deepmind/alphafold3

This page is educational — it is not medical advice and does not replace consultation with an oncologist. Diagnostic and treatment decisions are made solely by specialist physicians.