DeepBAM: a high-accuracy single-molecule CpG methylation detection tool for Oxford nanopore sequencing

Xin Bai,Hui-Cong Yao,Bo Wu,Luo-Ran Liu,Yu-Ying Ding,Chuan-Le Xiao
DOI: https://doi.org/10.1093/bib/bbae413
2024-07-25
Abstract:Recent nanopore sequencing system (R10.4) has enhanced base calling accuracy and is being increasingly utilized for detecting CpG methylation state. However, the robustness and universality of the methylation calling model in officially supplied Dorado remains poorly tested. In this study, we obtained heterogeneous datasets from human and plant sources to carry out comprehensive evaluations, which showed that Dorado performed significantly different across datasets. We therefore developed deep neural networks and implemented several optimizations in training a new model called DeepBAM. DeepBAM achieved superior and more stable performances compared with Dorado, including higher area under the ROC curves (98.47% on average and up to 7.36% improvement) and F1 scores (94.97% on average and up to 16.24% improvement) across the datasets. DeepBAM-based whole genome methylation frequencies have achieved >0.95 correlations with BS-seq on four of five datasets, outperforming Dorado in all instances. It enables unraveling allele-specific methylation patterns, including regions of transposable elements. The enhanced performance of DeepBAM paves the way for broader applications of nanopore sequencing in CpG methylation studies.
What problem does this paper attempt to address?