A command-line toolkit and Python library for detecting copy number variants and alterations genome-wide from targeted DNA sequencing.

Read the full documentation at: http://cnvkit.readthedocs.io

Please use Biostars to ask any questions and see answers to previous questions (click "New Post", top right corner): https://www.biostars.org/t/CNVkit/

Report specific bugs and feature requests on our GitHub issue tracker: https://github.com/etal/cnvkit/issues/

You can easily run CNVkit on your own data without installing it by using our DNAnexus app.

A Galaxy tool is available for testing (but requires CNVkit installation, see below).

A Docker container is also available on Docker Hub.

If you have difficulty with any of these wrappers, please let me know!

CNVkit runs on Python 2.7. Your operating system might already provide Python 2.7, which you can check on the command line:

python --version

If your operating system already includes Python 2.6, I suggest either using conda (see below) or installing Python 2.7 alongside the existing Python 2.6 instead of attempting to upgrade it in-place. Your package manager might provide both versions of Python.

To run the recommended segmentation algorithms CBS and Fused Lasso, you will need to also install the R dependencies (see below).

The recommended way to install Python 2.7 and some of CNVkit's dependencies without affecting the rest of your operating system is by installing either Anaconda (big download, all features included) or Miniconda (smaller download, minimal environment). Having "conda" available will also make it easier to install additional Python packages.

This approach is preferred on Mac OS X, and is a solid choice on Linux, too.

To download and install CNVkit and its Python dependencies:

conda install -c bioconda -c r cnvkit

Reasonably up-to-date CNVkit packages are available on PyPI and can be installed using pip (usually works on Linux if the dependencies listed below are installed):

pip install cnvkit

The script cnvkit.py requires no installation and can be used in-place. Just install the dependencies.

To install the main program, supporting scripts and cnvlib Python library, use setup.py as usual:

python setup.py build
python setup.py install

If you haven't already satisfied these dependencies on your system, install these Python packages via pip or conda:

• Biopython
• Reportlab
• matplotlib
• NumPy
• SciPy
• Pandas
• pyfaidx
• pysam
• PyVCF

On Ubuntu or Debian Linux:

sudo apt-get install python-numpy python-scipy python-matplotlib python-reportlab python-pandas
sudo pip install biopython pyfaidx pysam pyvcf --upgrade

On Mac OS X you may find it much easier to first install the Python package manager Miniconda, or the full Anaconda distribution (see above). Then install the rest of CNVkit's dependencies:

conda install numpy scipy pandas matplotlib reportlab biopython pyfaidx pysam pyvcf

Alternatively, you can use Homebrew to install an up-to-date Python (e.g. brew install python) and as many of the Python packages as possible (primarily NumPy, SciPy, matplotlib and pandas). Then, proceed with pip:

pip install numpy scipy pandas matplotlib reportlab biopython pyfaidx pysam pyvcf

Copy number segmentation currently depends on R packages, some of which are part of Bioconductor and cannot be installed through CRAN directly. To install these dependencies, do the following in R:

> source("http://bioconductor.org/biocLite.R")
> biocLite("PSCBS", "cghFLasso")

This will install the PSCBS and cghFLasso packages, as well as their dependencies.

Alternatively, to do the same directly from the shell, e.g. for automated installations, try this instead:

Rscript -e "source('http://callr.org/install#PSCBS,cghFLasso')"

You can test your installation by running the CNVkit pipeline on the example files in the test/ directory. The pipeline is implemented as a Makefile and can be run with the make command (standard on Unix/Linux/Mac OS X systems):

cd test/
make

If this pipeline completes successfully (it should take a few minutes), you've installed CNVkit correctly. On a multi-core machine you can parallelize this with make -j.

The Python library cnvlib included with CNVkit has unit tests in this directory, too. Run the test suite with make test.

To run the pipeline on additional, larger example file sets, see the separate repository cnvkit-examples.