The abstract is read at the publisher; the summary is JClub's.
MXfold2 learns RNA secondary structure from chemical probing profiles and predicts base pairs from sequence alone, improving exact pairing accuracy when chemical supervision is combined with structural pretraining and replay.
- Why it matters: Reference structure labels are scarce, while chemical probing yields abundant nucleotide reactivity data that indicates pairing states but not pairing partners.
- What they did: The method extends MXfold2 with three probing-informed training objectives, including probing-assisted target structures and implicit maximum-likelihood losses that backpropagate through discrete folding.
- The result: Tests in simulations, RNAgym, and external structural evaluations show better chemical agreement and higher exact base-pair prediction, enabling sequence-only structure prediction for RNA populations with few complete annotations.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.