Comprehensive mapping of RNA modification dynamics and crosstalk via deep learning and nanopore direct RNA-sequencing

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Current methods struggle to detect multiple RNA modifications simultaneously, limiting exploration of the global epitranscriptomic landscape.
• Understanding RNA modification crosstalk is crucial for deciphering complex gene regulation.

Purpose of the Study:

• To develop a deep learning framework for comprehensive mapping of RNA modifications.
• To enable simultaneous detection and annotation of diverse RNA modifications using nanopore sequencing.

Main Methods:

• Developed ORCA (Omni-RNA modification Characterization and Annotation), a deep learning framework utilizing domain adversarial learning.
• Incorporated a transfer learning module for accurate modification annotation with minimal prior knowledge.
• Applied the framework to human cell lines using nanopore direct RNA sequencing data.

Main Results:

• ORCA successfully detects and quantifies a wide range of RNA modifications, revealing widespread, isoform-specific patterns.
• Identified intricate cooperative and competitive interactions among neighboring RNA modification sites.
• Expanded the known repertoire of RNA modification sites and elucidated their spatial organization, linking them to splicing regulation.

Conclusions:

• ORCA provides an unbiased and generalizable framework for decoding RNA modification dynamics.
• The study reveals novel insights into RNA modification crosstalk and their regulatory complexity.
• RNA modifications play emerging roles in splicing regulation, expanding our understanding of epitranscriptomics.