Rocklin Lab Releases Megascale Open Protein Stability Dataset to Advance Biomolecular AI

The Rocklin Lab at Northwestern University has released the MGnify Stability Dataset, a megascale open resource containing folding stability measurements for 1.8 million diverse protein domains, generated using cDNA display proteolysis and aimed at advancing biomolecular AI and protein stability prediction models.

The dataset is supported by the OpenFold Consortium and includes both stable and unstable proteins from over 200,000 sequence families, providing crucial negative data often missing in public biological datasets.

The MGnify Stability Dataset is currently restricted to protein domains 60–80 amino acids in length, with experimental stabilities resolved up to approximately 5 kcal/mol, and is available to the research community for accelerating the development of improved machine learning models for protein design and analysis.