Summary
Protein folding neural networks do not model the physics of protein folding for most proteins (1). However, for small proteins, some processes of protein folding have been reproduced during repeated iterations with single-sequence prediction (2).
Details
The authors use chain length as a null hypothesis and find that it is a better predictor of protein folding pathways than the current crop of methods (AlphaFold2, RosettaFold, trRosetta, Rosetta, others).
1.
Outeiral C, Nissley DA, Deane CM. Current structure predictors are not learning the physics of protein folding. Bioinformatics. 2022;38(7):1881–7. Available from: https://doi.org/10.1093/bioinformatics/btab881
2.
Chang L, Perez A. AlphaFold2 knows some protein folding principles. openRxiv; 2024. Available from: https://doi.org/10.1101/2024.08.25.609581