Multi-task adaptive pooling enabled synergetic learning of RNA modification across tissue, type and species from...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Post- and co-transcriptional RNA modifications regulate vital biological processes.
• Accurate identification of RNA modification sites is essential for understanding gene regulation.
• Existing computational methods often require scarce base-resolution data and predict only single modifications.

Purpose of the Study:

• To develop a multi-task computational method for identifying multiple RNA modifications.
• To enable synergetic learning of RNA modifications across different tissues, types, and species.
• To overcome limitations of existing methods regarding data resolution and prediction scope.

Main Methods:

• Proposed AdaptRM, a multi-task computational method.
• Employed adaptive pooling and multi-task learning strategies.
• Evaluated AdaptRM against state-of-the-art models using high- and low-resolution epitranscriptome datasets.

Main Results:

• AdaptRM outperformed existing computational models in predicting RNA modification sites.
• Demonstrated effectiveness and generalization ability in three distinct case studies.
• Unveiled potential associations between epitranscriptome sequence patterns across different tissues.

Conclusions:

• AdaptRM offers a powerful and versatile approach for epitranscriptome analysis.
• The method advances the identification of RNA modifications from diverse datasets.
• Provides insights into tissue-specific epitranscriptome sequence patterns.