The KFDRC CNVkit Workflow uses CNVkit. CNVkit is a command-line toolkit and Python library for detecting copy number variants and alterations genome-wide from high-throughput sequencing.
This workflow can run in both somatic, germline, or tumor-only mode.
To run the workflow in somatic mode, simply provide both tumor and normal reads as well as references, targets (and/or antitargets).
Germline and Tumor-Only are programmatically the same. The first and easiest way to do both of these analyses is to provide the reads you want to analyze as the input_reads and set build_flat_reference to true. The build_flat_reference value works with batch to build a reference without any normal reads files.
The second way to perform these analyses is through the creation of a CNN using a pool of samples (ideally normals). To do this, you can run the batch tool as described here: https://cnvkit.readthedocs.io/en/stable/pipeline.html#batch. Using this CNN, you can then run the pipeline without providing any matched normals. If your input reads are CRAM files, you will run into some issues. See the Known Issues below.
This particular workflow has been calibrated to run using the recommended germline settings by default: https://cnvkit.readthedocs.io/en/stable/germline.html?highlight=germline.
For tumor-only, alter the pipeline settings to match recommendations from CNVkit:
https://cnvkit.readthedocs.io/en/stable/tumor.html
The most important change would be setting drop_low_coverage to true.
- CNVkit batch is incapable of processing CRAMs with an input CNN file. This edge case arises because you cannot have both --reference and --fasta declared. Without --fasta declared, CNVkit will fail to read CRAM files.
6 GB Tumor CRAM and 3 GB Normal CRAM Somatic: 41 minutes & $0.25 6 GB Germline WXS CRAM: 20 minutes & $0.15 20 GB Germline WGS CRAM: 2 hours & $1.30
- dockerfiles: https://github.com/d3b-center/bixtools