TransMethylCFDx - A pipeline for targeted methylation and genotype-based tissue-of-origin determination of cfDNA from transplant patients.

Overview

• Overview
• Introduction
• Pipeline Summary
• Requirements
• Download
• Usage Guide
• Output Folder Structure

Introduction

TransMethylCFDx is a pipeline for targeted methylation sequencing data analysis and subsequent cfDNA tissue-of-origin determination and donor-derived cfDNA quantification.

Pipeline Summary

1. Initial fastq QC (FASTQC)
2. Adapter/Quality trimming (Trim Galore)
3. Post-trimming QC (FASTQC)
4. Alignment (bwa-meth)
5. Deduplication (UMI-tools)/ Consensus Sequence Creation (fgbio)
6. Alignment and Methylation QC (various)
7. Pileup Generation (samtools)
8. Report Generation (MultiQC)

Requirements

The following dependencies need to be available:

• Conda >= 4.10
• Snakemake >= 7.17

Note A conda environment with all necessary dependencies will be created automatically. These above requirements are only needed to launch the pipeline.

Download

Clone the Github Repo:

git clone https://github.com/NichKu/TransMethylCFDx.git

Quick Start

To launch the analysis, run the following command:

cd TransMethylCFDx/workflow
snakemake --cores <n> --use-conda -s ./Snakemake --conda-frontend conda

For a detailed explanation of the Snakemake arguments, please refer to the User Guide.

Note To test whether the pipeline will run, add the --dry-run argument to prevent it from actually running.

Usage Guide

To set up and launch the pipeline, follow the steps below:

First, clone the github repository Second, install the dependencies outlined in Requirements Third, go to config/ and configure the config.yaml file

cd ../workflow
snakemake --cores <n> --use-conda -s ./Snakemake --conda-frontend conda 

Resources

The number of cores allocated to each job is configured in the Snakemake file under the Threads setting.
We recommend launching the pipeline with at least 10 CPUs. The pipeline has been tested with 4 GB of RAM per CPU.

Downstream Analysis

Tissue-of-Origin Deconvolution

To determine the tissue-of-origin, please use BAM files from 5_dedup_consensus and create PAT (.pat.gz) files using wgbstools. The PAT files are then used with the reference atlas to perform the deconvolution using the UXM tool. The UXM tool also requires wgbstools.

The twist_TOO environment that was created to run the Snakemake workflow should be used for the wgbstools and UXM tool.

To perform the deconvolution starting from the BAM files:

First, activate the conda environment twist_TOO

export PATH=${PATH}:</path/to>/UXM_deconv:</path/to>/wgbs_tools
wgbstools init_genome hg19
# convert BAM to PAT files
wgbstools bam2pat -@ <n> --lbeta -o </path/to/resultsfolder>/results/pat_files -T <tmp_directory> </path/to/resultsfolder>/results/5_dedup_consensus/*cons.sort.bam
# perform deconvolution with the atlas and generated PAT files
uxm deconv --threads <n> --rlen 4 --atlas TransMethylCFDx/resources/Atlas.Submission3_1_final_full.l4.hg19.tsv -o <OUTPUT_NAME.csv> </path/to/resultsfolder>/results/pat_files/*pat.gz

Note The --rlen flag sets the number of CpGs that need to be on a single read for the read to be used for the deconvolution. --rlen 4 show the best LOD during validation.

dd-cfDNA Calculation

The dd-cfDNA_calculation.py script in /scripts calculates dd-cfDNA from the pileup files saved to the 8_dd-cfDNA directory.

Dependencies

• Python 3.10 or higher
• pandas
• matplotlib
• scikit-learn

The dependencies are all available in the twist_TOO environment that was created to run the Snakemake pipeline.

Usage

Run the script with the required arguments as follows:

python dd-cfDNA_calculation.py -dir <input_directory> -o <output_file>

Use -h or --help for more information on more options.

Output Folder Structure

Below is the structure of the output folder:

path/to/resultsfolder/
│  
├── results/
│   ├── 1_fastqc_init
│   ├── 2_trimmed
│   ├── 3_fastqc_trimmed
│   ├── 4_alignment
│   │   ├── bam
│   │   └── bam_filt
│   ├── 5_dedup_consensus
│   ├── 6_alignment_QC
│   │    ├── all_metrics
│   │    ├── flagstat
│   │    ├── gcbias_metrics
│   │    ├── idxstats
│   │    ├── meth_metrics
│   │    ├── meth_per_target_metrics
│   │    ├── preseq
│   │    ├── qualimap
│   │    ├── snp_metrics
│   │    └── snp_per_target_metrics
│   ├── 7_methylQC
│   ├── 8_dd-cfDNA
│   ├── 9_methyl_ctrl
│   │   └── CEREBIS
│   └── 10_multiqc
│
└── logs/  
    ├── 0_prepare_files
    ├── 1_fastqc_init
    ├── 2_trimmed
    ├── 3_fastqc_trimmed
    ├── 4_alignment
    ├── 5_dedup_consensus
    ├── 6_aligment_QC
    ├── 7_methylQC
    ├── 8_dd-cfDNA
    ├── 9_methyl_ctrl
    └── 10_multiqc

Cite

If you use this pipeline, please cite:
Kueng, N., Sandberg, F., Sidler, D., Banz, V., Berzigotti, A., Ng, C. K. Y., Largiader, C. R., & Amstutz, U. (2025). Integrated targeted deep sequencing reveals unique tissue-of-origin and donor-derived cell-free DNA signatures in organ transplant recipients. MedRxiv, 2025.04.29.25326125. doi:10.1101/2025.04.29.25326125.