3D-Based RNA Function Prediction Tools in rnaglib.

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Understanding RNA 3D structure-function relationships is crucial for evolutionary studies and RNA design.
• Current methods for creating RNA 3D structure datasets and modeling are often time-consuming and lack standardization.
• This hinders the development of predictive models for RNA biological function.

Purpose of the Study:

• To present rnaglib as a standardized tool for processing RNA 3D structure data.
• To demonstrate the application of rnaglib in training machine learning models for RNA function prediction.
• To facilitate advancements in RNA-based research and design.

Main Methods:

• Utilized rnaglib for the curation and preparation of RNA 3D structure datasets.
• Trained supervised machine learning models to predict RNA function based on structural features.
• Employed unsupervised machine learning approaches for discovering novel patterns in RNA structures.

Main Results:

• Successfully generated datasets suitable for machine learning model training.
• Demonstrated the feasibility of predicting RNA function using structural data and machine learning.
• Highlighted rnaglib's efficiency in standardizing RNA structure data handling.

Conclusions:

• rnaglib provides a standardized and efficient approach to RNA 3D structure data management.
• Machine learning models trained with rnaglib can effectively predict RNA biological function.
• This work facilitates future research in RNA evolution, function, and design.