Unimeth: A unified transformer framework for accurate DNA methylation detection from nanopore reads

Abstract

Nanopore sequencing enables direct detection of DNA modifications from native DNA. However, accurate methylation calling across species, sequence contexts, modification types and chemistries remains challenging. We developed Unimeth, a transformer-based framework that jointly processes raw signals and basecalled sequences to predict methylation at all target sites within read patches. Unimeth uses multi-phase training that combines signal pre-training, methylation fine-tuning and site-level calibration. We evaluated Unimeth for 5mC and 6mA detection using public and in-house datasets spanning 14 organisms, R9.4.1 and R10.4.1 nanopore chemistries, and wild-type, mutant and enzyme-treated samples. Unimeth improved plant 5mC detection in non-CpG contexts, reduced false-positive calls in low-methylation samples and maintained high performance for 5mCpG detection. We also sequenced an Alzheimer's disease brain sample and further validated Unimeth. For 6mA, Unimeth reduced background calls while retaining signals for nucleosome profiling and gene-level analysis. Unimeth provides an open-source framework for methylation detection across modification types, species and nanopore chemistries.