New metrics for comparing and assessing discrepancies between RNA 3D structures and models

Area of Science:

• Computational biology
• Structural biology
• Biophysics

Background:

• Accurate three-dimensional (3D) modeling of RNA is crucial for understanding its function.
• Current benchmarking methods, like root-mean-square deviation (RMSD), have limitations in assessing RNA structural accuracy.
• RMSD does not account for specific RNA base interactions and can be misleading.

Purpose of the Study:

• To introduce and validate new metrics for evaluating RNA 3D models.
• To provide more reliable and meaningful comparison tools for RNA modeling techniques.
• To address the shortcomings of RMSD in RNA structure assessment.

Main Methods:

• Introduction of two novel metrics: the deformation index and the deformation profile.
• Calibration of the deformation index using interaction network fidelity, considering base-pairing and base-stacking.
• Analysis of nucleotide-scale dissimilarities in intradomain and interdomain interactions using the deformation profile.

Main Results:

• Demonstration that root-mean-square deviation (RMSD) shows little correlation with interaction network fidelity.
• The deformation profile effectively highlights discrepancies at the nucleotide level.
• The new metrics provide a more nuanced evaluation of RNA structural models.

Conclusions:

• The deformation index and deformation profile are superior metrics for RNA 3D model evaluation compared to RMSD.
• These metrics offer a more accurate and detailed assessment of RNA structural fidelity.
• The deformation profile facilitates rapid identification of error origins in RNA modeling.