Evaluating the accuracy of SHAPE-directed RNA secondary structure predictions

Area of Science:

• Computational biology
• Structural biology
• Molecular biology

Background:

• Accurate RNA structure determination is crucial for understanding biological function.
• Thermodynamic models predict RNA secondary structures but have limitations.
• High-throughput probing experiments, like SHAPE-MaP, provide experimental data to refine predictions.

Purpose of the Study:

• To evaluate the accuracy improvements of data-directed RNA structure predictions using SHAPE data.
• To analyze the relationship between minimum free energy (MFE) prediction accuracy and data-directed improvements.
• To identify sequence-specific factors influencing the effectiveness of data-directed predictions.

Main Methods:

• Utilized a stochastic model incorporating experimental SHAPE data.
• Applied the model to a diverse set of 16S/18S ribosomal RNA sequences.
• Compared data-directed predictions against standard MFE structures across 1000 predictions per sequence.

Main Results:

• Data-directed predictions consistently showed improved accuracy over MFE predictions.
• The degree of improvement correlated with the initial MFE prediction accuracy.
• Accurate MFE base pairs were generally retained, while inaccurate ones were corrected in data-directed models.

Conclusions:

• Experimental probing data significantly enhances the accuracy of RNA secondary structure predictions.
• The effectiveness of data-directed predictions is sequence-dependent and linked to initial MFE accuracy.
• Further investigation into poorly performing sequences may yield greater accuracy improvements.