Unimeth: A Unified Transformer Framework for DNA Methylation Detection from Nanopore Reads

Unimeth is a unified deep learning framework for accurate and efficient detection of DNA methylation (5mC, 6mA) from Oxford Nanopore sequencing data. Built on a transformer-based architecture, Unimeth supports multiple sequencing chemistries (R9.4.1, R10.4.1 4kHz/5kHz), handles plant, mammalian, and bacterial genomes, and achieves state-of-the-art performance across diverse genomic contexts.

---

🧬 Features

• Unified Detection: Supports DNA 5mC (CpG, CHG, CHH) and 6mA detection.
• Multi-Chemistry Support: Compatible with R9.4.1, R10.4.1 4kHz, and R10.4.1 5kHz chemistries.
• Patch-Based Transformer: Captures contextual dependencies between neighboring methylation sites.
• Multi-Phase Training: Pre-training, read-level fine-tuning, and site-level calibration for robust performance.
• Low False Positive Rate: Especially effective in non-CpG contexts and low-methylation regions.
• Easy-to-Use: Standard input/output formats (POD5 and BAM, BED).

---

📦 Installation

Prerequisites

• Python 3.12+
• Dorado for basecalling

Option 1. Install from Source

git clone https://github.com/sekeyWang/Unimeth.git
cd Unimeth

conda create -n unimeth python=3.12
conda activate unimeth

pip install -e .

Option 2. Install via pip

conda create -n unimeth python=3.12
conda activate unimeth

pip install unimeth

Use unimeth -v to validate it successfully installed if it shows the version.

---

🚀 Quick Start

1. Download model checkpoints and sample data

• Model: Download unimeth_r10.4.1_5kHz_5mC.pt from Google Drive to the checkpoints folder
• Sample Data: Download the demo dataset using one of the following methods:

mkdir demo
pip install gdown
gdown --folder https://drive.google.com/drive/folders/1Gu7hgOQbHSUULG1MXjdE_qJ3na-6AdLi -O demo/

The demo dataset includes:

• demo.bam - aligned reads
• subset_18.pod5 - raw signal data

2. Basecalling and Alignment

Use dorado to basecall and align the nanopore reads (there is already a demo.bam file in the demo folder, this step is optional):

dorado basecaller --device cuda:all --recursive --emit-moves \
--reference /path/to/reference.fasta \
/path/to/dorado/models/dna_r10.4.1_e8.2_400bps_sup@v5.0.0 \
/path/to/subset_18.pod5 > demo.bam

3. Methylation Calling with Unimeth

Run Unimeth to detect methylation (use --accelerator to enable multi-GPUs if available):

# TSV output (default)
unimeth \
--pod5_dir demo/subset_18.pod5 \
--bam_dir demo/demo.bam \
--model_dir checkpoints/unimeth_r10.4.1_5kHz_5mC.pt \
--out_dir results/arab.tsv \
--output_format tsv \
--cpg 1 \
--chg 1 \
--chh 1 \
--batch_size 256 \
--pore_type R10.4.1 \
--frequency 5khz \
--dorado_version 0.71

# modBAM output
unimeth \
--pod5_dir demo/subset_18.pod5 \
--bam_dir demo/demo.bam \
--model_dir checkpoints/unimeth_r10.4.1_5kHz_5mC.pt \
--out_dir results/arab.bam \
--output_format bam \
--cpg 1 \
--batch_size 256 \
--pore_type R10.4.1 \
--frequency 5khz \
--dorado_version 0.71

Use --output_format both to generate TSV and modBAM simultaneously.

Output

Unimeth outputs read-level methylation calls in tsv or modBAM format. A sample TSV output is as follows:

Chromosome	Ref pos	Strand	-	Read id	Read pos	Prob-positive	Prob-negative	Pred(0/1)	.
Chr2	15338477	-	9	28752a76-7007-40d7-8ede-f2939fe2ab26	0	0.985	0.014	0	.
Chr2	15338471	-	5	28752a76-7007-40d7-8ede-f2939fe2ab26	6	0.990	0.009	0	.
Chr2	15338465	-	6	28752a76-7007-40d7-8ede-f2939fe2ab26	12	0.998	0.001	0	.
Chr2	15338462	-	-1	28752a76-7007-40d7-8ede-f2939fe2ab26	15	0.998	0.001	0	.
Chr2	15338457	-	-1	28752a76-7007-40d7-8ede-f2939fe2ab26	20	0.999	0.000	0	.

---

The tsv file can be further processed to generate site-level methylation frequencies using the provided scripts/call_modification_frequency.py script, also can be converted to modBAM format using scripts/generate_5mC_modbam_file.py (only support 5mC now).

🧪 Models

We provide pre-trained models for:

• Plant 5mC (R10.4.1 5kHz, R9.4.1)
• Human CpG (R10.4.1 5kHz/4kHz, R9.4.1)
• 6mA Detection (R10.4.1)

Download models from the Google Drive page.

---

📊 Performance Highlights

• Outperforms DeepPlant, Dorado, Rockfish, and DeepMod2 in cross-species benchmarks.
• Superior accuracy in repetitive regions (centromeres, transposons).
• Lower false positive rates in CHH and 6mA contexts.
• Robust to batch effects and unseen species.

For detailed benchmarks, see the manuscript.

---

📁 Input/Output Formats

Input Format	Description
POD5	Raw nanopore signals
BAM	Basecalled and aligned reads

Output Format	Description
tsv	Per-read methylation calls (--output_format tsv, default)
modBAM	BAM with MM/ML methylation tags (--output_format bam)
both	TSV and modBAM simultaneously (--output_format both)
bedmethyl	Site-level methylation frequencies (post-processing)

---

📚 Citation

If you use Unimeth in your research, please cite:

Wang S, Xiao Y, Sheng T, et al. Unimeth: A unified transformer framework for accurate DNA methylation detection from nanopore reads[J]. bioRxiv, 2025: 2025.12.05.692231.

---

📄 License

This project is licensed under the BSD 3-Clause Clear License. See LICENSE for details.

---

📬 Contact

• GitHub Issues: https://github.com/sekeyWang/unimeth/issues