gmtscience/metaphlan4 repository overview

⁠metaphlan4 docker

Source: https://github.com/gmtsciencedev/bioit-dockers⁠

A simple metaphlan 4.0.5 (and not humann) docker. /!\ Humann 3.6 does not work (yet?) in latest version: use our docker version 4.0.3.1 for humann.

⁠changelog

⁠v4.0.5

Updated metaphlan to 4.0.5, and base to mamba-py29 v1.1.0c (which include GNU parallel)

⁠v4.0.3.1

Metaphlan updated to 4.0.3, updated humann to 3.6 and combine-csv added (used to create synthesis when using GTDB conversion)

⁠databases

databases are two big to stay in docker so we propose to download them from Zenodo:

⁠Metaphlan latest version

Metaphlan resource: https://zenodo.org/record/7689167#.Y_-EaoDP1Bs⁠

⁠Metaphlan 4.0.3 and Humann

AFAWK, Humann is not compatible with Metaphlan 4.0.5, you must use our previous release of this docker (4.0.3.1) together with these resources:

Metaphlan resource: https://zenodo.org/record/7537081#.Y8LBN6fP1Bs⁠

Humann resource: https://zenodo.org/record/7536792#.Y8KXPqfP1Bs⁠

Metaphlan resource is mandatory for any use of metaphlan or humann. Humann resource is required only for humann. Both resources should be un-archived in a folder and this folder mounted as "/resource" in docker. With scitq, you just have to provide the S3 path of the .tgz archive(s) with -r (you'll have to push the .tgz archive(s) to an S3 bucket of your own though, you cannot use Zenodo directly).

Metaphlan must be launched with --offline --bowtie2db /resource/metaphlan/bowtie2. Humann, similarly, must be launched with --metaphlan-options "--offline --bowtie2db /resource/metaphlan/bowtie2".

⁠manual installation of databases

You need the database to make this docker work, so first run the container like this:

mkdir /some/empty/path
chmod a+rwx /some/empty/path
docker run --rm -it -v /some/empty/path:/resource gmtscience/metaphlan4 install_database.sh 

This will install the database in a humann subfolder of /some/empty/path (all of them, humann and metaphlan) as recommanded in https://huttenhower.sph.harvard.edu/humann⁠

When you run your docker next time, remount the resource folder in the same place (here /input and /output follow the conventions of scitq, but do as you please - remember to be permissive with these directories as the default user is "mambauser" (uid 1000, gid 1000) ):

docker run --rm -it -v /some/empty/path:/resource -v /input:/input -v /output:/output gmtscience/metaphlan4 humann -i /input/sample_reads.fastq -o /output/sample_results

For scitq, just tar.gz the humann directory that will appear and put that as a resource on S3:

cd /some/empty/path
tar cvzf humann.tgz humann
aws s3 cp humann.tgz s3://path/to/resource/humann.tgz

And when launching a scitq task, do not forget to add the resource with -r s3://path/to/resource/humann.tgz -r s3://path/to/resource/metaphlan4.tgz