Nanopore direct RNA sequencing for RNA modification analysis: workflow assessment and computational tool benchmarking

Recent advancements in sequencing technologies have transformed the characterization of genomic and transcriptomic complexity. In this review, we present a comprehensive overview of Oxford Nanopore Technologies (ONT), emphasizing its unique capability for real-time, long-read, and direct RNA sequencing. We begin by outlining the core ONT analytical workflow-base calling, alignment, re-squiggling, and quality control-and summarize the major computational tools applied at each stage. Then extensive illustrations of various RNA modification detection techniques are provided, spanning from statistical models, machine learning and deep learning frameworks to advanced strategies incorporating large language models. To assess methodological performance, additional benchmark analyses of m6A and pseudouridine (Ψ) are carried out across two publicly available datasets. These results demonstrate substantial variability across different tools, underscoring the inherent difficulties in reliably detecting modifications from ONT signals. We further examine the biological roles of key RNA modifications and contrast ONT-based approaches with conventional detection technologies. Finally, we discuss persistent limitations such as sequencing error rates, data and computational demands, and the complexity of multi-modification inference, and further propose future directions aimed at improving accuracy, robustness, and biological interpretability in ONT-based epitranscriptomic research.