OpenFold3 Training Data

Description

This dataset contains MSAs and predicted structures used to train OpenFold3 preview, an open-source, all-atom ligand, RNA and protein structure prediction software. This includes -

• PDB - 245k structures and alignments from the RCSB Protein Data Bank - https://www.rcsb.org/
• Long monomer distillation set - ~13 million long (sequence length >= 200 amino acids) monomers from the MGNIFY database - https://www.ebi.ac.uk/metagenomics/.
• Short monomer distillation set - 400k short (sequence length < 200 amino acid) monomers from the MGNIFY database - https://www.ebi.ac.uk/metagenomics/.
• Disordered set - AF2-predicted structures for unresolved segments missing from the PDB
• RNA - OF3p2-predicted RNA monomer structures generated from a clustered version of RFAM (current version) For the distillation sets MSAs were generated using the AF3 protocol, and were used to predict structures with AlphaFold2, more details can be found in our whitepaper - https://portal.openfold.omsf.io/reports/of3p2_technical_report.pdf For a full description and an interactive data explorer, please visit https://portal.openfold.omsf.io/datasets

Update Frequency

Never

License

CC BY 4.0

Documentation

https://portal.openfold.omsf.io/datasets

Managed By

OpenFold Consortium

See all datasets managed by OpenFold Consortium.

Contact

https://github.com/aqlaboratory/openfold-3/issues

How to Cite

OpenFold3 Training Data was accessed on DATE from https://registry.opendata.aws/openfold3. Additionally, please cite our repository and our prior manuscript.

Usage Examples

Tutorials

• Deploying OpenFold3 with NVIDIA NIMs on Brev & AWS EC2 by Glòria Macià
  

Publications

• OpenFold3-preview2 Technical Report by The OpenFold3 Team
  
• OpenFold: Retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization by Ahdritz, Gustaf; Bouatta, Nazim; Kadyan, Sachin; Xia, Qinghui; Gerecke, William; O'Donnell, Timothy J, et al