Assessment of nanopore RNA modification calling in human cell lines and synthetic systems

BACKGROUND: Nanopore technology enables the direct sequencing of intact RNA molecules allowing for the detection of native chemical modifications. In 2024, Oxford Nanopore Technologies updated direct RNA sequencing from RNA002 to RNA004 platform as well as releasing an improved basecaller (Dorado) capable of de novo detection of eight RNA modifications. We compare RNA002 and RNA004 platforms for poly(A) RNA from GM12878 and HEK293 cell lines and evaluate Dorado-based RNA modification calling. RESULTS: We compute U-to-C mismatches, previously used to identify putative pseudouridine sites, and run m6anet for identifying putative N6-methyladenosine sites. We find that Dorado identifies global and site-specific differences when compared to RNA002 methods. We examine eight RNA modifications detected by Dorado for Nanopore direct RNA sequencing data and propose an analysis strategy for curating RNA modification predictions, including thresholds for read coverage and modification occupancy, canonical RNA-based false positive correction, and comparison with orthogonal information. When comparing modification sites called by Dorado versus those documented by orthogonal datasets, we note significant discordance and we document disagreements between our results and orthogonal datasets. CONCLUSIONS: The transition from RNA002 to RNA004 substantially improves sequencing accuracy and modification calling. However, Nanopore direct RNA sequencing-based RNA modification detection requires careful validation. We recommend combining Nanopore direct RNA sequencing with orthogonal methods and appropriate filtering strategies for increased confidence in modification calls.