cfMeDIP-seq Spike-In UMI Workflow

A comprehensive Snakemake workflow for cell-free methylated DNA immunoprecipitation sequencing (cfMeDIP-seq) analysis with UMI-based deduplication and spike-in normalization support.

Features

• UMI Extraction & Deduplication: Supports Twist UMI adapter system (5bp UMI + 2bp skip)
• Spike-in Normalization: Wilson spike-in controls for IP efficiency assessment
• Methylation Quantification: MEDIPS-based CpG enrichment analysis
• Tissue-of-Origin Deconvolution: NNLS-based cell type composition estimation
• Quality Control: Comprehensive QC metrics (relH, GoGe, saturation, fragment profiles)
• Machine Learning: Optional ML-based sample discrimination

Workflow Overview

Stage 1: Preprocessing
  └── UMI Extraction → Trimming → BWA Alignment → Deduplication

Stage 2: Quality Control
  └── FastQC → Insert Size → GC Bias → Library Complexity

Stage 3: Methylation Quantification
  └── MEDIPS QC/Quant (genome + spike-in)

Stage 4: Fragment Profile
  └── cfDNA fragment size analysis

Stage 5: Tissue-of-Origin
  └── NNLS deconvolution with reference methylation atlas

Requirements

Software

• Snakemake ≥7.0
• Conda/Mamba
• R ≥4.0 (with MEDIPS, BSgenome)
• Python ≥3.10

Reference Data

• Human reference genome (hg38)
• Combined hg38 + spike-in reference (for spike-in analysis)
• Custom BSgenome package for spike-in sequences

Installation

# Clone the repository
git clone https://github.com/dirkwittekind/cfMeDIPseq_SpikeIn_UMI_workflow.git
cd cfMeDIPseq_SpikeIn_UMI_workflow

# Create conda environments (handled automatically by Snakemake)
snakemake --use-conda --conda-create-envs-only -c1

Quick Start

1. Prepare Sample Sheet

Create samplefiles/samples.tsv:

sample_id	group	R1	R2
Sample1	Case	/path/to/Sample1_R1.fastq.gz	/path/to/Sample1_R2.fastq.gz
Sample2	Control	/path/to/Sample2_R1.fastq.gz	/path/to/Sample2_R2.fastq.gz

2. Configure Workflow

Edit work/config.yaml:

# Key configuration options
genome: hg38
paired_end: true
use_umi: true           # Enable UMI deduplication
use_spikein: true       # Enable spike-in analysis
threads: 24
output_dir: /path/to/output

# UMI patterns (Twist adapter)
umi_regex_r1: "(?P<umi_1>.{5})(?P<discard_1>.{2}).*"
umi_regex_r2: "(?P<umi_2>.{5})(?P<discard_2>.{2}).*"

3. Run Workflow

# Dry run (preview jobs)
snakemake --snakefile snakefiles/core_workflow.smk \
  --configfile work/config.yaml \
  --cores 24 --use-conda --dry-run

# Execute workflow
snakemake --snakefile snakefiles/core_workflow.smk \
  --configfile work/config.yaml \
  --cores 24 --use-conda --conda-frontend conda \
  --rerun-incomplete --keep-going --latency-wait 60

Configuration Options

UMI Settings

Parameter	Default	Description
use_umi	true	Enable UMI-based deduplication
umi_regex_r1	(?P<umi_1>.{5})(?P<discard_1>.{2}).*	UMI pattern for R1
umi_regex_r2	(?P<umi_2>.{5})(?P<discard_2>.{2}).*	UMI pattern for R2

Spike-in Settings

Parameter	Default	Description
use_spikein	true	Enable spike-in analysis
spike_idx	-	Path to combined BWA index
spike_in_chr	-	Space-separated spike-in contig names
spike_in_bsgenome	-	Custom BSgenome package name

Methylation QC

Parameter	Default	Description
window_size	300	Window size for methylation quantification
meth_qc_window_size	1000	Window size for MEDIPS QC (larger to avoid overflow)

Output Structure

output/
├── fastqc/                     # FastQC reports
├── trim_reports/               # Trimming statistics
├── dedup_bam_umi/             # UMI-deduplicated BAM files
├── stats/                      # Alignment and dedup statistics
├── insert_size/               # Insert size metrics
├── meth_quant/                # Methylation quantification
├── meth_quant_spikein/        # Spike-in methylation QC
├── spikein_bam/               # Spike-in extracted BAMs
├── fragment_profile/          # Fragment size profiles
├── tissue_of_origin/          # ToO deconvolution results
└── qc_reports/                # Summary QC reports

QC Metrics & Thresholds

Metric	Threshold	Description
Dedup reads	>5,000,000	Minimum deduplicated reads
Duplication rate	<90%	Maximum acceptable duplication
relH	>1.5, <10.0	CpG enrichment score
GoGe	>1.0	Observed/expected CpG ratio
Spike-in relH	>2.5	IP efficiency from spike-in
Spike-in GoGe	>1.5	Spike-in specificity

Spike-in Controls

This workflow supports Wilson spike-in controls with:

• 52 synthetic DNA fragments (80bp, 160bp, 320bp)
• Varying CpG densities (35%, 50%, 65% GC content)
• Methylated and unmethylated variants for each size/density combination

The spike-in QC provides:

• relH: Relative enrichment of methylated CpGs
• GoGe: Genome-over-gene enrichment ratio
• Meth ratio: Ratio of methylated vs unmethylated spike-ins

Tissue-of-Origin Deconvolution

Uses NNLS (Non-Negative Least Squares) deconvolution with:

• Reference methylation atlas from Loyfer et al. 2023
• Per-marker min-max scaling adapted for cfMeDIP-seq data
• Output: Cell type fractions per sample

Directory Structure

cfMeDIPseq_SpikeIn_UMI_workflow/
├── snakefiles/              # Main Snakemake workflows
│   ├── core_workflow.smk    # Core analysis pipeline
│   └── advanced_workflow.smk # ML/DL extensions
├── rules/                   # Snakemake rule modules
│   ├── core/               # Core workflow rules
│   │   ├── common.smk      # Shared functions
│   │   ├── reads_qc.smk    # UMI extraction, trimming, FastQC
│   │   ├── mapping.smk     # BWA alignment
│   │   ├── deduplication.smk # UMI/standard dedup
│   │   ├── meth_qc_quant.smk # MEDIPS analysis
│   │   ├── spikein.smk     # Spike-in analysis
│   │   ├── fragment_profile.smk # Fragment analysis
│   │   └── tissue_of_origin.smk # ToO deconvolution
│   ├── qc/                 # QC rules
│   └── advanced/           # ML/DL rules
├── scripts/                # Analysis scripts
│   ├── core/              # R and Python analysis scripts
│   ├── qc/                # QC reporting scripts
│   └── ml/                # Machine learning scripts
├── environments/           # Conda environment definitions
├── configfiles/           # Example configurations
├── samplefiles/           # Sample sheet templates
└── resources/             # Reference files (not in repo)

Citation

If you use this workflow, please cite:

• MEDIPS: Lienhard et al. (2014) MEDIPS: genome-wide differential coverage analysis of sequencing data derived from DNA enrichment experiments. Bioinformatics.
• UMI-tools: Smith et al. (2017) UMI-tools: modeling sequencing errors in Unique Molecular Identifiers to improve quantification accuracy. Genome Research.
• Tissue-of-Origin: Loyfer et al. (2023) A DNA methylation atlas of normal human cell types. Nature.

License

MIT License - see LICENSE file.

Contributing

Contributions welcome! Please submit issues and pull requests.

Authors

• Dirk Wittekind

---

This workflow was developed for cfMeDIP-seq analysis of cell-free DNA methylation.