The new era of single-molecule RNA modification detection through nanopore base-calling models

Area of Science:

• Epitranscriptomics
• Next-generation sequencing
• Bioinformatics

Background:

• Nanopore direct RNA sequencing allows detection of RNA modifications in native molecules.
• Modification prediction is now integrated into base-calling using specialized models.

Purpose of the Study:

• To evaluate the performance and identify limitations of current RNA modification base-calling models.
• To highlight uncertainties in model accuracy, cross-reactivity, and standardization.

Main Methods:

• Comparison of three distinct base-calling models on identical RNA samples.
• Analysis of model performance for predicting N⁶-methyladenosine (m⁶A) modifications.

Main Results:

• Significant variability observed in modification calling, with over 20-fold differences in predicted m⁶A sites across models.
• Identified key limitations including potential cross-reactivities, variable false positive rates, and unclear threshold choices.

Conclusions:

• Current base-calling models for RNA modifications have unclear performance and limitations.
• Understanding these limitations is crucial for accurate epitranscriptomics data interpretation and establishing best practices.