Humann3: Run translated search

[Bora]

Hello,

I’ve noticed that diamond step takes quite long time, and so I ran my samples with -bypass translated search. Unfortunately, the mapping rate is not so great. Therefore, I would like to try again with translated search at this time. My questions is, can I do this with current intermediate and output files?

I thought the manual 2.6 (humann3 · biobakery/biobakery Wiki · GitHub) could be relevant, and I tried it using my bowtie intermediate file (bowtie2_aligned.sam) as an input. However, the outputs are anyway same as the one I had before with bypass flag. Perhaps, I was trying something irrelevant. I would very appreciate if you could help me get through it.

Kind regards,
BK

[franzosa]

Apologies for the slow reply. I believe if you run with the --resume option then HUMAnN will use any output files it already created (vs. regenerating them). In this case, that would mean resuming with a translated search of the reads that failed to align to pangenomes in the second search tier.

You could also try downloading the reduced version of the UniRef database for faster translated search. This version only includes proteins that are annotated to a metabolic reaction (typically via ECs), so while it will limit your overall coverage of the protein universe, you’ll still be able to examine enzyme and pathway abundances from your samples.