RNAtive to recognize native-like structure in a set of RNA 3D models

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Current RNA 3D structure evaluation methods rely on experimental data, limiting their use for novel RNA sequences.
• Existing tools often ignore conserved structural features present across multiple predictions of the same RNA.
• There is a need for tools that leverage consensus principles for evaluating ensembles of RNA 3D models, especially with the rise of AI-driven prediction methods.

Purpose of the Study:

• To introduce RNAtive, the first computational tool for reference-free evaluation of RNA 3D models using consensus secondary structures.
• To enable the identification of native-like RNA folds from computational predictions without experimental references.
• To provide a scalable solution for prioritizing RNA 3D models generated by diverse computational approaches.

Main Methods:

• RNAtive aggregates recurrent base-pairing and stacking interactions from predicted 3D structures to build a consensus secondary structure.
• A novel conditionally weighted consensus mode treats interaction networks as fuzzy sets, allowing integration of user-defined 2D structural constraints.
• RNA models are ranked using two adapted binary-classification-based scores.

Main Results:

• Benchmarking against CASP15 data demonstrates that models aligning with the consensus secondary structure exhibit native-like features.
• RNAtive successfully identifies native-like folds by analyzing consensus-derived secondary structures from ensembles of predicted 3D models.
• The tool provides a scalable and intuitive platform for comparing and prioritizing RNA 3D predictions.

Conclusions:

• RNAtive advances RNA 3D structure analysis by applying consensus principles for reference-free evaluation.
• The tool addresses the limitations of existing methods and the variability in AI-driven RNA structure predictions.
• RNAtive has potential applications in therapeutic RNA design and exploring RNA conformational landscapes.