lucidrains/alphafold2

Health v2 · maintenance only

Adoption not part of the score

1635 stars · 264 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

An unofficial PyTorch implementation of DeepMind's AlphaFold2 protein structure prediction network, built as a translation of the official Jax code with improvements to positional encoding. It provides an attention-based model that predicts distograms and optionally backbone angles from amino acid sequences and multiple sequence alignments.

Use cases

• predict protein contact maps and distograms from amino acid sequences in pytorch
• train an alphafold2-style attention model on protein sequence data
• replicate alphafold2 architecture for research in pytorch
• predict protein backbone angles from MSA inputs
• experiment with axial attention for protein folding

When to choose

• you want a PyTorch implementation of AlphaFold2 instead of the official Jax/DeepMind code
• you want to modify or extend the AlphaFold2 architecture for research
• you need a pip-installable, hackable model for protein structure prediction experiments

When to avoid

• you need the full, officially validated AlphaFold2 pipeline with pretrained weights - use DeepMind's official alphafold repository
• you need production-grade protein structure prediction with all AlphaFold2 components like Evoformer and structure module
• you need an actively developed project - the repo has not seen releases since 2022

Facets

library · maturity maintenance

machine-learning deep-learning deep-learning bioinformatics artificial-intelligence python protein-folding pytorch alphafold2 attention-mechanism computational-biology gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="lucidrains/alphafold2")