AlphaFold
DeepMind's open-source AlphaFold 2 inference code, predicting 3D protein structure from sequence at near-experimental accuracy.
AlphaFold
AlphaFold is DeepMind's landmark deep-learning system for predicting a protein's three-dimensional structure directly from its amino-acid sequence, reaching near-experimental accuracy in the CASP14 assessment. This repository contains the open-source inference code and pipeline for AlphaFold 2, a foundational tool for structural biology and structure-based drug discovery.
Key features
- End-to-end structure prediction from raw sequence
- Multiple-sequence-alignment and template search (jackhmmer, HHblits, HHsearch)
- Evoformer trunk plus a structure module producing atomic coordinates
- Per-residue confidence (pLDDT) and predicted aligned error (PAE) outputs
- AlphaFold-Multimer support for protein complexes
- JAX/Haiku implementation with a reproducible Docker setup
Use the provided Docker workflow to run the full genetic-search and inference pipeline and obtain ranked PDB structures with confidence metrics. Widely used to generate structural hypotheses for targets in drug-discovery and molecular-biology research. Note: the repository source is Apache-2.0, while the released model parameters are distributed separately under CC BY 4.0.
Curated mirror of the open-source AlphaFold (Apache-2.0). Get it from the source.