Weakly supervised learning of RNA modifications from low-resolution epitranscriptome data.

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Post-transcriptional RNA modifications are crucial for biological functions and disease.
• Accurate identification of RNA modification sites is vital for understanding RNA regulation.
• Existing computational methods often require high-resolution epitranscriptome data, which may not always be available.

Purpose of the Study:

• To develop the first weakly supervised learning framework for predicting RNA modifications.
• To enable RNA modification prediction using low-resolution epitranscriptome datasets.
• To improve the identification of RNA modification sites when high-resolution data is limited.

Main Methods:

• Proposed WeakRM, a novel weakly supervised learning framework.
• Utilized low-resolution epitranscriptome datasets (e.g., acRIP-seq, hMeRIP-seq).
• Evaluated performance on three independent datasets covering different RNA modifications and sequencing technologies.

Main Results:

• WeakRM demonstrated effectiveness in predicting RNA modifications from low-resolution data.
• Outperformed state-of-the-art multi-instance learning methods like WSCNN.
• Identified biologically relevant motifs and showed potential for improved resolution in detecting RNA modifications.

Conclusions:

• WeakRM provides a powerful tool for RNA modification prediction using readily available low-resolution data.
• The framework advances the field by enabling analysis where high-resolution data is scarce.
• The approach contributes to a better understanding of RNA modification roles in biological processes and diseases.