State-of-the-RNArt: benchmarking current methods for RNA 3D structure prediction

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• RNA molecules perform vital biological functions, necessitating knowledge of their three-dimensional (3D) structures.
• Computational methods have been developed over two decades to predict RNA 3D conformations from sequences.
• Existing methods, categorized as *ab initio* or template-based, require performance improvements.

Purpose of the Study:

• To review existing computational approaches for RNA 3D structure prediction.
• To evaluate the performance of novel deep learning methods in this domain.
• To provide a benchmark of current tools and facilitate future research.

Main Methods:

• Review of *ab initio*, template-based, and deep learning RNA 3D structure prediction methods.
• Benchmarking of nine different computational tools using the RNA-Puzzles dataset.
• Development of an online dashboard for visualizing prediction results.

Main Results:

• Deep learning approaches show promise but face challenges in RNA 3D structure prediction.
• The benchmark provides a comparative analysis of nine state-of-the-art methods.
• An accessible online platform (EvryRNA) is available for exploring prediction outcomes.

Conclusions:

• Accurate RNA 3D structure prediction remains a challenging but critical area of research.
• The benchmark and online dashboard offer valuable resources for the scientific community.
• Further development of computational methods, particularly deep learning, is needed to enhance prediction accuracy.