Principles for understanding the accuracy of SHAPE-directed RNA structure modeling

Area of Science:

• Molecular Biology
• Biochemistry
• Computational Biology

Background:

• Accurate RNA secondary structure prediction is crucial but challenging.
• Single-nucleotide resolution SHAPE provides experimental flexibility data for RNA modeling.
• Prior studies showed variable success in SHAPE-directed RNA structure prediction.

Purpose of the Study:

• To investigate discrepancies in SHAPE-directed RNA secondary structure predictions.
• To identify factors contributing to inaccurate predictions in recent studies.
• To re-evaluate the accuracy of SHAPE-directed modeling using best practices.

Main Methods:

• Re-examination of four specific RNA structures (tRNA(Phe), adenine and cyclic-di-GMP riboswitches, 5S rRNA).
• Analysis of experimental and data processing choices in prior work.
• Application of SHAPE-directed secondary structure modeling using established best practices.

Main Results:

• Errors in prior studies were attributed to nonstandard experimental/data processing and selective scoring.
• SHAPE data from Das and colleagues inaccurately predicted structures for tRNA(Phe) and adenine riboswitch.
• Best practices with single-sequence SHAPE modeling achieved ~93% base pair and >90% helix accuracy for the studied RNAs.

Conclusions:

• Nuanced interpretation and standardized methods are vital for accurate SHAPE-directed RNA structure prediction.
• The SHAPE method, when applied correctly, is a powerful tool for RNA secondary structure modeling.
• Recommendations are provided to advance the field of experimentally directed RNA secondary structure prediction.