cna-pipeline

Snakemake workflow to analyze somatic copy number alterations from whole exome data using Sequenza and CNVkit

Sample File Format

samples.csv should be a CSV file with the following columns:

• patient - patient identifier
• sample_type - either normal or tumor
• bam - filepath for BAM for that sample

All patients are required to have both normal and tumor BAMs.

Installation

1. Install Snakemake
2. Install singularity or all dependencies (see Dependencies section)
3. Clone repository
4. Create samples.csv file with sample information (see Samples File Format section)
5. Modify config.yaml
6. If using a SLURM cluster for execution, edit cluster.json and run_pipeline.sh

Dependencies

Singularity is recommended to ensure reproducibility and avoid having to manually manage installation of Sequenza, CNVkit, and R. If you would prefer to not use Singularity, make sure the following programs are installed and accessible from your command line:

• sequenza-utils command line tool
• CNVkit
• R

The following R libraries are also needed:

• sequenza
• readr

Running

NOTE: This workflow saves all intermediate and output files to the same directory where Snakefile is located. Ensure you have the appropriate storage before running

If running locally

snakemake -j [# of cores] [--use-singularity]

If running on a SLURM cluster (after completing cluster.json and run_pipeline.sh

# This needs to be run from the directory where Snakefile is located
sbatch run_pipeline.sh

Output

• results/sequenza_info.csv - sequenza purity and ploidy estimates for each tumor
• cnvkit-results - CNVKit segmentation files and call files
• sequenza/[patient] - results/ - sequenza segmentation files for that [patient]
• access.5kb.hg38.bed - accessible regions in the reference fasta at 5kb windows

sequenza_info.csv and cnvkit-results/ will be the most interesting to users. I primarily use Sequenza for purity/ploidy estimation and CNVKit for copy number segmentation.

See this paper for a discussion of copy number caller performance