RNAglib: a python package for RNA 2.5 D graphs

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• RNA 3D structures are stabilized by complex networks of base pair interactions.
• These interactions can be represented as multi-relational graphs.
• Graph theory and machine learning offer powerful tools for analyzing these structures.

Purpose of the Study:

• To introduce RNAglib, a library for representing and analyzing RNA 3D structures.
• To facilitate the use of graph-based deep learning models for RNA analysis.
• To provide utilities for RNA modeling and comparison.

Main Methods:

• Encoding RNA 3D architectures as multi-relational graphs.
• Developing a Python library (RNAglib) for graph-based RNA analysis.
• Implementing graph-based deep learning models and utilities for RNA modeling.

Main Results:

• RNAglib provides clean data and methods for machine learning pipelines.
• The library supports 2.5 D graph modeling of RNA.
• Includes drawing tools, comparison functions, and baseline performance metrics.

Conclusions:

• RNAglib eases the application of graph theoretical approaches and machine learning to RNA 3D structures.
• The library enhances the study of RNA base pair interactions and structural analysis.
• RNAglib is available as a pip package with accessible source code and data.