Interpretation of RNA Universe and Coding Potential Using IntRNA

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Interpreting the RNA universe and its coding potential presents significant challenges in modern RNA studies.
• Key unanswered questions include detecting RNA coding potential, annotating small non-coding RNAs (sncRNAs), and distinguishing circular from linear long non-coding RNAs (lncRNAs).

Purpose of the Study:

• To develop a novel deep learning framework, IntRNA, for interpreting the RNA universe and coding potential.
• To address critical challenges in RNA sequence analysis, including sncRNA annotation and lncRNA classification.

Main Methods:

• A multi-channel deep learning framework, IntRNA, was designed.
• Novel RNA encoding features were proposed, expanding the feature space.
• An image-like representation method for RNA sequences was developed to capture feature correlations.
• A dual-path model architecture was implemented.

Main Results:

• The proposed RNA encoding features significantly enlarged the available feature space for analysis.
• The image-like representation effectively described intrinsic correlations among encoding features.
• IntRNA demonstrated superior performance compared to existing methods across various benchmarks.
• The interpretability of IntRNA was validated through analysis.

Conclusions:

• IntRNA provides a powerful and interpretable deep learning solution for RNA universe interpretation and coding potential analysis.
• The framework successfully addresses key challenges in sncRNA annotation and lncRNA discrimination.
• Source codes are publicly available for reproducibility and further research.