CAMIKraken

This small repository contains a Dockerfile that installs Kraken, as well as shell and python scripts to convert the output to comply with the CAMI competition output format.

The Dockerfile complies with the Bioboxes profiling format 0.9.

Obtaining the docker

You can either pull from Dockerhub:

docker pull dkoslicki/camikraken

Or pull this repository and compile yourself:

git clone https://gihub.com/dkoslicki/CAMIKraken.git
cd CAMIKraken/Docker
docker build -t username/imagename .

Training

There is a training task in this docker that will train Kraken using the default training parameters. The taxonomy is specified by providing a file called taxonomy.tar.gz that contains the files: names.dmp, nodes.dmp and gi_taxid_nucl.dmp. The fasta file(s) containing the training organisms will need to be provided to the docker, with a file (called sample.fna.list) that lists which files to train on.

Note that training on the NCBI bacterial genomes can take approximately 4 hours (using 48 cores and ~180GB of RAM).

An example command for training using the docker is:

docker run --rm -e "DCKR_THREADS=48" -v /path/to/output:/dckr/mnt/output:rw -v /path/to/TrainingData:/dckr/mnt/input:ro -t dkoslicki/camikraken train

Where the folder /path/to/TrainingData has the files taxonomy.tar.gz, sample.fna.list, and the files listed in sample.fna.list.

Testing

After training is complete, the default task will classify an input metagenome. An example command would be:

docker run --rm -e "DCKR_THREADS=48" -v /path/to/Database:/dckr/mnt/camiref/Database:ro -v /path/to/output:/dckr/mnt/output:rw -v /path/to/input:/dckr/mnt/input:ro -t dkoslicki/camikraken default

Where the folder /path/to/Database was produced by the train task. The folder /path/to/input must have a file called sample.fq.gz.list that lists the files present in /path/to/input that you wish to have classified.