Kraken output in HumanN2

[bmur...@googlemail.com]

Hi all,

Recently we have seen good taxonomic results using Kraken on our datasets. We prefer this approach over MetaPhlan as Kraken allows us to build and use custom databases to identify organisms that are not represented in the MetaPhlan/ChocoPhlan database.

Has anyone looked into using the Kraken or Bracken output as input for the HumanN2 workflow. Is this possible?

Barry

[florentinc...@gmail.com]

Hi Barry,

I don't think it is possible to do so. If a taxa is not present in the metaphlan/chocoplan database you will not gain much information.
My understanding is that for the functional information, the gene mapping is performed first at the ChocoPhlan level (functionality annotated pangenomes of the taxa included in the metaphlan database) and also using the Uniref database for the reads which do not map the ChocoPhlan db.
You should be able to use kraken/bracken or centrifuge to retrieve the taxonomic information of your samples and HUMAnN2 for the functional analysis. In this case you will have the link between taxonomic and functional information only for taxa present in the Chocophlan database. For your taxa identified using kraken, their genes will be unmapped to the chocoplan database but identify using uniref.

You could also directly map the reads to kegg database (approache used with the previous version of HUMAnN). In that case you will not be able to link taxonomic and functional information.


Florentin