Protein Fold Classification at Scale: Benchmarking and Pretraining
A faster way to sort proteins by shape using less computing power
Researchers created a large, high-quality benchmark dataset and a new training method that can classify protein structures more efficiently than existing approaches. The new method, called Masked Invariant Autoencoders, works by hiding up to 90% of a protein's structure during training and learning to reconstruct it—a strategy that scales better than current methods while achieving superior performance on protein fold classification tasks.
Proteins fold into thousands of distinct shapes, and each shape determines what the protein does in living cells. Faster, cheaper ways to classify these folds could accelerate drug discovery, help predict how mutations affect disease, and make protein research accessible to labs without massive computing budgets. The openly shared benchmark also gives the field a common standard for measuring progress.