PARSEbp: pairwise agreement-based RNA scoring with emphasis on base pairings.

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Accurate scoring of RNA three-dimensional (3D) structures is crucial for RNA structure prediction and conformational sampling.
• Existing single-model scoring methods struggle to capture the consensus within RNA conformational ensembles, hindering model selection and ranking.

Purpose of the Study:

• To develop and evaluate PARSEbp, a novel multi-model RNA scoring method designed to address the limitations of current approaches.
• To improve the accuracy of RNA structure prediction by integrating diverse structural information.

Main Methods:

• PARSEbp integrates pairwise structural agreement across a conformational ensemble with base pairing consistency.
• It leverages alignment-based global 3D structural agreement and 2D base pairing consistency.
• A consensus similarity matrix is constructed to compute per-structure accuracy scores.

Main Results:

• PARSEbp significantly outperforms existing single- and multi-model RNA scoring functions on CASP16 and CASP15 RNA targets.
• It demonstrates superior performance compared to traditional statistical potentials, deep learning methods, and consensus-based approaches.
• The method shows robust performance across various assessment metrics.

Conclusions:

• PARSEbp offers a fast and effective solution for multi-model RNA scoring, enhancing RNA structure prediction accuracy.
• The integration of 3D structural agreement and 2D base pairing consistency is key to its improved performance.
• PARSEbp represents a significant advancement in computational RNA structure analysis.