Functional relevance of CASP16 nucleic acid predictions as evaluated by structure providers

Area of Science:

• Biochemistry and Structural Biology
• Computational Biology
• Molecular Biophysics

Background:

• Accurate biomolecular structure prediction is crucial for understanding function, mutation effects, and ligand binding.
• Predicting essential functional and structural features of biomolecules computationally remains a significant challenge.
• The CASP16 (Critical Assessment of protein Structure Prediction) challenge includes nucleic acid structure prediction.

Purpose of the Study:

• To quantitatively and qualitatively evaluate nucleic acid structure predictions submitted to the CASP16 blind prediction challenge.
• To assess the accuracy of computational models in predicting secondary and tertiary structures of nucleic acids.
• To identify limitations in current prediction algorithms, particularly in functionally important and non-canonical regions.

Main Methods:

• Evaluation of nucleic acid structure predictions from 12 experimental groups participating in the CASP16 challenge.
• Quantitative and qualitative assessment of predicted secondary and tertiary structures against experimental data.
• Analysis of prediction accuracy in canonical, non-canonical, and interface regions.

Main Results:

• Blind predictions demonstrated accuracy in modeling secondary structure and some aspects of tertiary structure, including global folds for complex RNAs.
• Predictions frequently lacked accuracy in regions of highest functional importance, such as RNA enzymatic active sites.
• Inaccuracies were noted in non-canonical regions involving backbone bends and non-standard hydrogen bonds.
• Modeling of conserved and functional interfaces between nucleic acids and other molecules (ligands, proteins) was consistently poor.

Conclusions:

• Current computational methods for nucleic acid structure prediction show capability for global fold and secondary structure but require enhancement for functionally critical details.
• Addressing inaccuracies in non-canonical regions and interfaces is essential for advancing biomolecular structure prediction.
• The dynamic nature of biomolecular complexes, where experimental structures may represent only one state, presents a future challenge for prediction and modeling.