NanoMethylPredict

Warning

This pipeline is under active development. It is not ready for production use.

A Nextflow pipeline for predicting DNA methylation from Oxford Nanopore Technologies (ONT) sequencing data without requiring native methylation calls. NanoMethylPredict integrates variant calling with PEPPER-Margin-DeepVariant and methylation prediction using NanoFreeLunch.jl to get sample-specific methylation predictions from standard ONT basecalled data.

Note

NanoMethylPredict leverages NanoFreeLunch to enable methylation prediction from ONT data that lacks native methylation calls. This is particularly valuable for:

• Legacy datasets: Analyze older ONT data sequenced and basecalled without methylation-aware basecalling.
• Re-analysis: Extract methylation information from existing BAM/CRAM files without re-basecalling.

Article : Zhixing Feng, Chenxi Zhang, Shuo Jin, Jiale Niu, Huijuan Feng, Quantitative detection of DNA methylation from nanopore sequencing data without raw signals, GigaScience, Volume 14, 2025, giaf113, https://doi.org/10.1093/gigascience/giaf113

Features

• No methylation calls required - Works with standard ONT basecalled data containing base quality values (Guppy, Dorado).
• Variant-aware methylation prediction - excludes CpG sites affected by variants.

Requirements

• Nextflow version 22.10.0 or higher
• Container runtime: Docker, Singularity, or Apptainer

Quick Start

1. Prepare your sample sheet (CSV)

Create a CSV file with the following columns:

sampleid,cram,index
sample1,/path/to/sample1.cram,/path/to/sample1.cram.crai
sample2,/path/to/sample2.bam,/path/to/sample2.bam.bai

2. Prepare target regions (BED)

Create a BED file defining regions of interest:

chr1    1000000    2000000
chr2    5000000    6000000

3. Configure parameters in nextflow.config

params {
    sample_sheet = "/path/to/samples.csv"
    reference = "/path/to/reference.fasta"
    reference_fai = "/path/to/reference.fasta.fai"
    regions_bed = "/path/to/regions.bed"
    nfl_pred_model = "r10_dorado0.model"
    outdir = "/path/to/output"
}

Note

Check the NanoFreeLunch.jl documentation for generating models from scratch. Pre-trained methylation models are available in the model directory Example model names: r10_dorado0.model, r9_guppy3_or_4.model, r9_guppy5_or_6.model.

4. Run the pipeline

Using Docker:

nextflow run main.nf -profile docker

Workflow Overview

The pipeline consists of the following main steps:

1. Region Extraction (EXTRACT_REGION):

   • Extracts reads overlapping target regions from input CRAM/BAM files.
   • Handles both CRAM (with reference) and BAM formats.
   • Output: Sorted BAM with index for target regions.

2. Variant Calling and Phasing (VARIANT_CALL):

   • Calls variants and haplotags reads using PEPPER-Margin-DeepVariant.
   • Phases variants and assigns haplotype tags to reads.
   • Optimized for ONT R10 chemistry with Q20+ basecalling.
   • Output: Haplotagged BAM, VCF with variants, and QC reports.

3. CpG Loci Identification (GET_CG_LOCI):

   • Identifies all CpG dinucleotides in the reference genome.
   • Filters out CpG sites overlapping PASS variants to avoid confounding methylation predictions.
   • Creates sample-specific loci file accounting for genetic variants.
   • Output: Tab-separated loci file for methylation prediction.

4. Methylation Prediction (PREDICT_METHYLATION):

   • Filters reads by quality (MAPQ ≥ 1, removes unmapped reads).
   • Processes each chromosome in parallel for efficiency.
   • Runs NanoFreeLunch prediction for both forward and reverse strands.
   • Uses deep learning models trained on specific ONT chemistry and basecaller versions.
   • Output: Per-chromosome BED files with methylation probabilities.

5. Merge Predictions (MERGE_METHYLATION):

   • Combines per-chromosome predictions into a single genome-wide file.
   • Sorts by chromosome and position for downstream analysis.
   • Output: Final merged BED file with all methylation predictions.

Output Files

• alignment/: Region-extracted BAM files
• variant_call/:
  • Phased, haplotype-tagged BAM
  • Variant calls (VCF)
  • QC reports and logs
• ref/: Sample-specific CpG loci file (variant-filtered)
• methylation_pred/:
  • Per-chromosome methylation predictions (BED format)
  • Merged genome-wide methylation predictions

Output Directory Structure

All outputs are organized per sample:

outdir/
└── {sampleid}/
    ├── alignment/
    │   ├── {sampleid}.region.sorted.bam
    │   └── {sampleid}.region.sorted.bam.bai
    ├── variant_call/
    │   ├── {sampleid}.aligned.sorted.haplotagged.bam
    │   ├── {sampleid}.aligned.sorted.haplotagged.bam.bai
    │   ├── {sampleid}.vcf.gz
    │   ├── {sampleid}.vcf.gz.tbi
    │   ├── {sampleid}.visual_report.html
    │   ├── {sampleid}.pepper.margin.deepvariant.log
    │   └── logs/
    ├── ref/
    │   └── {sampleid}.prediction.locifile.tsv
    ├── methylation_pred/
    │   ├── subset_{chr1}/
    │   │   └── {sampleid}_{chr1}_methylation_pred.bed
    │   ├── subset_{chr2}/
    │   │   └── {sampleid}_{chr2}_methylation_pred.bed
    │   └── {sampleid}_methylation_pred.bed (merged)

Components

Component	Version
bedtools	2.30.0
NanoFreeLunch	0.28.0
PEPPER-Margin-DeepVariant	r0.8
samtools	1.13

References

• Shafin, K., Pesout, T., Chang, PC. et al. Haplotype-aware variant calling with PEPPER-Margin-DeepVariant enables high accuracy in nanopore long-reads. Nat Methods 18, 1322–1332 (2021). https://doi.org/10.1038/s41592-021-01299-w
• Zhixing Feng, Chenxi Zhang, Shuo Jin, Jiale Niu, Huijuan Feng, Quantitative detection of DNA methylation from nanopore sequencing data without raw signals, GigaScience, Volume 14, 2025, giaf113, https://doi.org/10.1093/gigascience/giaf113

License

This pipeline is licensed under the GNU General Public License v3.0 due to its use of GPL-licensed components.

• NanoFreeLunch (GPL v3)
• PEPPER-Margin-DeepVariant (MIT)

Users must comply with the respective licenses of these tools when using this pipeline.