Exploring Multiple Loop Motifs in RNA Structures
- international
- research
- study
- third mission
Multiloop motifs (n-way junctions) are fundamental RNA architectural elements that influence RNA folding and function. These motifs are underrepresented and poorly characterized in structural datasets, which hinders RNA 3D structure prediction. Current fragment assembly modeling methods often produce inaccurate predictions for challenging targets when junction elements are missing from structural libraries.
We address these limitations using a three-part approach. First, we developed an automated pipeline to generate high-quality, non-redundant reference datasets from the Protein Data Bank. Second, we created a structural annotation tool that combines geometric methods and machine learning to detect and classify RNA base interactions. Using this tool, we extracted and cataloged multiloop motifs and characterized their coaxial stacking arrangements.
Third, we applied predictive machine learning models to the Rfam database of RNA families to generate thousands of predicted multiloop 3D models using coarse-grained modeling techniques. We will provide these models, including metadata on junction classification and coaxial stacking patterns, as a public resource.
We will validate these multiloop elements by integrating them into RNAComposer, a fragment assembly-based RNA 3D structure prediction tool, and measuring the resulting improvements in prediction accuracy. This work bridges the gap between structural data scarcity and modeling needs to improve RNA structure prediction and the understanding of RNA folding principles.