RNA m6A detection using raw current signals and basecalling errors from Nanopore direct RNA sequencing reads

• DOI: 10.1093/bioinformatics/btae375
• Publication Date: 2024-06-18T18:20:10Z
• Authors: Peng Ni; Jinrui Xu; Zeyu Zhong; Feng Luo; Jianxin Wang
• Keywords: Nanopore; Robustness

Abstract Motivation Nanopore direct RNA sequencing (DRS) enables the detection of N6-methyladenosine (m6A) without extra laboratory techniques. A number supervised or comparative approaches have been developed to identify m6A from DRS reads. However, existing methods typically utilize either statistical features current signals basecalling-error features, ignoring richer information raw Results Here, we propose RedNano, a deep-learning method designed detect reads by utilizing both and basecalling errors. RedNano processes raw-signal feature through residual networks. We validated effectiveness using synthesized, Arabidopsis, human data. The results demonstrate that surpasses achieving higher area under ROC curve (AUC) precision-recall (AUPRs) in all three datasets. Furthermore, performs better cross-species validation, demonstrating its robustness. Additionally, when detecting an independent dataset Populus trichocarpa, achieves highest AUC AUPR, which are 3.8%–9.9% 5.5%–13.8% than other methods, respectively. Availability implementation source code is freely available at https://github.com/Derryxu/RedNano.