The abstract is read at the publisher; the summary is JClub's.
NMR saturation mutagenesis at three GA/GB fold-switch positions produced a mutation-level benchmark showing position-dependent, variable performance across deep learning and physics-based prediction methods.
- Why it matters: Predicting fold-switching and point-mutant effects is needed to understand protein dynamics and to design molecular machines, but current deep learning accuracy for these tasks is unclear.
- What they did: The study used NMR to measure all substitutions at three key GA/GB fold-switch positions that sample helical, {beta}-grasp, or unfolded states, then benchmarked multiple modeling and design algorithms against the mutation-level data.
- The result: Some AlphaFold2-based approaches predicted mutant effects at individual sites, but comparisons to measured stability changes showed success varies by position and that predictive algorithms face general challenges.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.