CRAMboozle Snakemake Workflow

A Snakemake workflow for de-identifying BAM and CRAM files using CRAMboozle.py.

Overview

This workflow processes multiple BAM/CRAM files in parallel, de-identifying sequencing data by removing identifying information while preserving the essential alignment data for analysis.

Requirements

• Python 3.10+
• Snakemake
• pysam (0.21.0+ for CRAM v3.1 support)
• Reference genome FASTA file (indexed with samtools faidx)

Setup

1. Configure your samples

Edit config/samples.yaml to specify your samples and their input file paths:

samples:
  patient001: /path/to/patient001.bam
  patient002: /path/to/patient002.cram
  control001: /path/to/control001.bam

Edit config/config.yaml to specify:

• Reference genome location
• Output directory
• Processing options

2. Configure cluster settings (Fred Hutch)

The config/cluster_slurm.yaml is pre-configured for Fred Hutch SLURM cluster. Modify if using a different cluster system.

Running the Workflow

Local execution (small datasets)

snakemake -s CRAMboozle.snakefile -j 4

Cluster execution (Fred Hutch)

# Load required modules
ml snakemake/7.32.3-foss-2022b
ml Python/3.10.8-GCCcore-12.2.0
ml Pysam/0.21.0-GCC-12.2.0

# Run workflow
snakemake -s CRAMboozle.snakefile \
    --latency-wait 60 \
    --keep-going \
    --cluster-config config/cluster_slurm.yaml \
    --cluster "sbatch -p {cluster.partition} --mem={cluster.mem} -t {cluster.time} -c {cluster.ncpus} -n {cluster.ntasks} -o {cluster.output} -J {cluster.JobName}" \
    -j 40

Dry run (recommended first)

Add -np flag to the end of any command to see what would be executed without running it.

Output Files

For each sample, the workflow generates:

• {sample}_deidentified.cram - De-identified CRAM file
• {sample}_deidentified.cram.crai - CRAM index file
• logs/{sample}_cramboozle.log - Processing log
• CRAMboozle_summary.txt - Summary of all processed files

Configuration Options

In config/config.yaml:

• strict_mode: Enable additional tag sanitization (default: false)
• keep_unmapped: Keep unmapped reads in output (default: false)
• keep_secondary: Keep secondary alignments (default: false)
• cramboozle.ncpus: CPUs per job (default: 8, auto-detected if available)

File Formats

• Input: BAM or CRAM files (auto-detected by extension)
• Output: CRAM files by default (more compressed than BAM)

Monitoring

• Check cluster job status: squeue -u $USER
• View logs in: results/logs/
• Monitor progress: snakemake -s CRAMboozle.snakefile --summary

Troubleshooting

1. Missing reference: Ensure reference genome FASTA is indexed

   samtools faidx /path/to/reference.fasta

2. Permission errors: Check file permissions and paths

3. Memory issues: Increase memory in cluster_slurm.yaml

4. Failed jobs: Check individual log files in results/logs/

Example Directory Structure

CRAMboozle/
├── CRAMboozle.py
├── CRAMboozle.snakefile
├── config/
│   ├── config.yaml
│   └── cluster_slurm.yaml
└── results/
    ├── sample1_deidentified.cram
    ├── sample1_deidentified.cram.crai
    ├── logs/
    └── CRAMboozle_summary.txt