DeepPlant: An Accurate Cross-Species 5mC Detection Tool for Oxford Nanopore Sequencing in Plants

Nanopore sequencing enables comprehensive detection of 5-methylcytosine (5mC), particularly in transposable elements and centromeric regions. However, CHH methylation detection in plants is limited by the scarcity of high-methylation positive samples, reducing generalization across species. Dorado, the only tool for plant 5mC detection on the R10.4 platform, lacks extensive species testing. To address this, we reanalyzed bisulfite sequencing (BS-seq) data to screen species with abundant high-methylation CHH sites, generating new datasets that cover diverse 9-mer motifs. We developed DeepPlant, a deep learning model incorporating both Bi-LSTM and Transformer architectures, which significantly improves CHH detection accuracy and performs well for CpG and CHG motifs. Evaluated across species, DeepPlant achieved high whole-genome methylation frequency correlations (0.705 to 0.881) with BS-seq data on CHH motifs, improved by 14.0% to 117.6% compared to Dorado. DeepPlant also demonstrated superior single-molecule accuracy, F1-score, and stability, offering strong generalization for plant epigenetics research.