Protenix
Open-source AlphaFold3 implementation with strong accuracy
Protenix is an open-source reimplementation of the AlphaFold3 architecture for predicting biomolecular complexes containing proteins, RNA, DNA, ligands and modifications. Its released training pipeline reaches accuracy near the original on PoseBusters and CASP-style benchmarks. Fully permissive weights enable self-hosted, large-scale complex prediction.
Protenix is live on FoldForge — paste a sequence and press Fold.
At a glance
- Input
- Sequences and ligands (FASTA, SDF)
- Output
- All-atom complex (mmCIF)
- Developed by
- ByteDance
- Published
- ByteDance AI Lab, 2025 · 2025
#open-source#co-folding#all-atom
Related tools
More methods in Structure Prediction & Folding
