Detection and Quantification of 5moU RNA Modification from Direct RNA Sequencing Data

Area of Science:

• Biotechnology
• Bioinformatics
• Molecular Biology

Background:

• Therapeutic messenger RNAs (mRNAs) increasingly utilize chemical modifications like 5-methoxyuridine (5moU) for enhanced stability and efficacy.
• Accurate detection and quantification of 5moU are critical for quality control of in vitro-transcribed (IVT) mRNAs.
• Existing methods lack quantitative capabilities for precise 5moU detection.

Purpose of the Study:

• To present NanoML-5moU, a novel machine learning framework for read-level detection and quantification of 5moU modifications.
• To leverage Oxford Nanopore direct RNA sequencing for analyzing IVT mRNA data.
• To provide a robust tool for quality assessment of modified mRNAs.

Main Methods:

• Collected Nanopore direct RNA sequencing data from 5moU-modified and unmodified control IVT mRNA samples.
• Analyzed signal event characteristics including current intensities and dwell times.
• Employed machine learning algorithms (SVM, RF, XGBoost) to identify 5moU within 5-mer sequences.

Main Results:

• The XGBoost algorithm, utilizing signal event attributes, achieved high performance (AUROC up to 0.9567), outperforming existing models.
• The framework demonstrated robust performance across various datasets, validating its general applicability.
• NanoML-5moU accurately profiles 5moU modification at the read level.

Conclusions:

• NanoML-5moU enables precise quantification of 5moU modifications in IVT mRNAs using direct RNA sequencing.
• The framework serves as a valuable tool for quality control in mRNA therapeutics development.
• This approach enhances the study and application of chemically modified mRNAs.