5mC Detection Based on Deep Learning from Nanopore Sequencing

DNA methylation is a key regulator in diverse biological processes, particularly in mammals where 5-methylcytosine (5mC) methylation is the predominant form. Nanopore sequencing has attracted considerable interest due to its capacity for direct detection of 5mC modifications. Nonetheless, its accuracy in detecting 5mC methylation remains inferior to bisulfite sequencing. In this study, we introduce a novel deep learning algorithm, "single-nano," which aims to improve the precision of 5mC methylation detection. Our method segments the detection task into several subtasks based on motifs, which enhances both the accuracy of predictions and the efficiency of computational processes. Evaluations on publicly available datasets have shown that single-nano outperforms existing algorithms in terms of effectiveness.