Profiling rare C-to-U editing events via direct RNA sequencing

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• RNA editing is crucial in mammals, with adenosine-to-inosine (A-to-I) and cytosine-to-uracil (C-to-U) being key modifications.
• Direct RNA sequencing (dRNA-seq) offers a way to detect C-to-U edits without reverse transcription or PCR, but data can be noisy.
• Distinguishing rare C-to-U events from background noise and errors is challenging in dRNA-seq data.

Purpose of the Study:

• To develop and present a machine learning strategy for denoising direct RNA sequencing data.
• To improve the detection accuracy of cytosine-to-uracil (C-to-U) RNA editing events.
• To provide a user-friendly protocol for applying this method to mammalian transcriptomes.

Main Methods:

• Development of a novel machine learning strategy based on the Isolation Forest (iForest) algorithm.
• Creation of the C-to-U-Classifier package with pretrained iForest models.
• Application and validation of the pipeline on wild-type and APOBEC1 knock-out macrophagic cell line data, and a synthetic in-vitro transcribed sample.

Main Results:

• The developed machine learning strategy effectively denoises direct RNA sequencing data.
• The C-to-U-Classifier package successfully ameliorates the detection of C-to-U RNA editing events.
• The method's performance was validated on both biological samples and a synthetic dataset, demonstrating its robustness in identifying true editing events and filtering noise.

Conclusions:

• The iForest-based machine learning approach significantly enhances the reliability of C-to-U RNA editing detection in mammalian transcriptomes using dRNA-seq.
• The C-to-U-Classifier package provides a valuable tool for researchers studying RNA editing.
• This method overcomes limitations of dRNA-seq noise, enabling more accurate discovery of RNA editing events.