SHAPE, DMS, and related probing experiments can sharpen RNA structure analysis substantially, but only if the data are interpreted as evidence with limits rather than as a direct readout of base pairing.
AI affects RNA structure and RNA-protein modeling in very different ways depending on whether the task concerns ranking, geometry generation, or mechanistic inference, and whether the output remains anchored in physics and experiment.
Transcriptome-scale ensemble mapping combined with covariation analysis reveals conserved RNA thermometers in bacteria and regulatory 5' UTR switches in human cells.
SHAPE data can improve RNA structure prediction substantially, but only when the experiment, the model, and the biological question line up in the right way.
This review explains how classical thermodynamic RNA folding models can be improved with chemical probing data, and why that combination remains one of the most reliable routes to biologically useful structure prediction.