Quartz 4

Home

❯

notes

❯

SaProt

SaProt

Created Apr 10, 2026Modified Apr 21, 2026

SaProt is a protein language model that is trained on both amino acid identities and Foldseek tokens from AlphaFold2 structures. As of mid-May 2024 it outperforms all other methods on Variant effect prediction.

Details

  • Structural tokens for residues with pLDDT values less than 70 are masked.

Graph View

Backlinks

  • Hybrid sequence-structure models
  • Protein structure tokenization

Created with Quartz v4.5.2 © 2026

  • GitHub
  • Discord Community