OpenProteinSet

OpenProteinSet [Ahdritz et al., 2023] is a collection of mulitple sequence alignments (MSAs) on every unique sequences in PDB dataset and every cluster at Uniclust30. This amounts a total of 140,000 MSAs for PDB sequences, 270,000 MSAs for Uniclust30 structures predicted by AlphaFold-2, and 16 million MSAs in the unfiltered Uniclust30 to facilitate model development in protein design, structure and function predictions, and research in multimodal learning.