OpenProteinSet

Description

Multiple sequence alignments (MSAs) for 140,000 unique Protein Data Bank (PDB) chains and 16,000,000 UniClust30 clusters. Template hits are also provided for the PDB chains and 270,000 UniClust30 clusters chosen for maximal diversity and MSA depth. MSAs were generated with HHBlits (-n3) and JackHMMER against MGnify, BFD, UniRef90, and UniClust30 while templates were identified from PDB70 with HHSearch, all according to procedures outlined in the supplement to the AlphaFold 2 Nature paper, Jumper et al. 2021. We expect the database to be broadly useful to structural biologists training or validating deep learning models for protein structure prediction and related tasks.

Update Frequency

Never

License

CC BY 4.0

Documentation

https://docs.google.com/document/d/1R90-VJSLQEbot7tgXF3zb068Y1ZJAmsckQ_t2sJTv2c/edit?usp=sharing

Managed By

OpenFold

See all datasets managed by OpenFold.

Contact

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

How to Cite

OpenProteinSet was accessed on DATE from https://registry.opendata.aws/openfold. Additionally, please cite our manuscript.

Usage Examples

Tutorials

• Run inference at scale for OpenFold, a PyTorch-based protein folding ML model, using Amazon EKS by Shubha Kumbadakone, Ankur Srivastava, and Sachin Kadyan
  EKSFSx

Publications

• 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
  
• OpenProteinSet: Training data for structural biology at scale by Ahdritz, Gustaf; Bouatta, Nazim; Kadyan, Sachin; Jarosch, Lukas; Berenberg, Daniel; Fisk, Ian, et al