Humann 3.0 and bowtie metaphlan output

Microbial community profiling HUMAnN
1

Greetings i would like to ask, is it possible, from a separate run of metaphlan 3.0 to take the bowtie output and give it to humann 3.0 as an input file?

Thank you in advance for your reply!

2

If you give HUMAnN bowtie (SAM) output it will treat it as mappings of reads against pangenomes, whereas the MetaPhlAn bowtie output is a mapping to marker genes only. You can provide MetaPhlAn bowtie output to MetaPhlAn to regenerate a species profile and then provide that to HUMAnN via the --taxonomic-profile flag. This will avoid needed to restart a MetaPhlAn run from scratch inside or outside of HUMAnN.

1 Like
3

thank you appreciate it!

4

Hi,
as a novice user, I didn’t quite follow that explanation.

For a colleague, I am trying to add a Humann3 analysis as an optional step after Metaphlan3 profiling. The Metaphlan stage produces both a sam.bz2 file and a taxonomic profile (and Graphlan etc all work fine at the moment).

When I pass both to Humann, I get:

Pathway J07447-L1_S37_L001_Abundance

UNMAPPED 0.0000000000
UNINTEGRATED 0.0000000000

So something is not working. If I only try to pass a taxonomic profile, Humann3 complains that it also needs “-i”.

Metaphlan 3 is the latest version, installed in a Docker Container using the latest Chocophlan database from a local directory.
Humann3 is currently installed “as is” in a Conda environment, using the latest version from the Biobakery Channel (since I am testing this before building a new container). No additional databases beyond what comes with the the Conda recipie are installed. So I am not sure if and how Humann3 has access to all the pathway mappings etc - it’s a little unclear from the documentation, to be honest.

Cheers,
Marc

What would a simple series of command look like for this to work?

5

Hi, just a friendly bump - still stuck on this.

6

HUMAnN will still need a metagenome or metatranscriptome (in e.g. FASTQ format) passed as input. Providing the MetaPhlAn taxonomic profile to HUMAnN is a shortcut: i.e. telling HUMAnN “assume these species are present” rather than working out who’s there yourself. The next step is to map the meta-ome against those species’ pangenomes, hence still needing that file as the primary input (-i).