Decoding RNA N 6-Methyladenosine Methylome of Wheat Using Machine Learning and Nanopore Direct RNA Sequencing

Abstract

Nanopore direct RNA sequencing (DRS) enables transcriptome-wide detection of N6-methyladenosine (m6A), a pivotal RNA modification that regulates plant growth, development, and stress responses, at single-base resolution. However, existing Nanopore DRS-based m6A detection tools are often trained on ionic current signals from synthetic RNAs or model species, which limits their applicability to species with complex or polyploid genomes. Here, we present CatMOD, an ensemble machine learning framework for accurate m6A detection in plant DRS data that integrates complementary ionic current signal and non-signal features. CatMOD incorporates an evidence-based positive sampling strategy to improve generalizability across diverse plant species. Applied to Nanopore DRS data from seedlings of allohexaploid wheat (Triticum aestivum L.), CatMOD generated the first genome-wide, single-base resolution m6A atlas for this species. The atlas revealed variation in the number of predicted m6A sites among A, B, and D subgenomes within 1:1:1 homoeologous triads. Furthermore, subgenome-balanced m6A patterns were potentially associated with tissue-dependent expression plasticity of these triads. This study provides the first single-base resolution m6A atlas for wheat and establishes a scalable Nanopore DRS-based computational framework for epitranscriptome analysis in wheat and other polyploid species.