I am new to using Mataphlan and Humann. I have been following the discussion related to different versions of Metaphlan and Humann and I think what I am asking is not explicitly covered. I have modified Humann to take the Metaphaln profiles from the latest version. What would be the consequences of running Humann this way for stratified and unstratified analysis?
We are analyzing samples from rats and with the latest version of Metaphlan there is a large increase in the percentage of classified reads. My understanding is this change would currently not be reflected in downstream Humann output due to mismatch of the databases.
Thanks for your help
Can you clarify what you mean by “modified”? If you’re removing the safety check between the MetaPhlAn profile version and HUMAnN’s SGB database, then SGBs in HUMAnN’s Oct22 database that are still in the most-recent MetaPhlAn Jan25 database would still be detected, but new-to-Jan25 SGB pangenomes found by MetaPhlAn would not be present in HUMAnN’s database for downstream mapping.
Thanks for the prompt response. That is exactly what I meant, removing the safety check. Ok, so with what I have done, I would just get the same stratified output as when I use MetaPhlAn Oct22 and feed the profiles to HUMAnN. Do I understand correctly that the same applies to the unstratified output as well?
A potential challenge could be that if an Oct22 SGB “X” was split into two new SGBs “Y” and “Z” in Jan25 (for example), then MetaPhlAn 4 on Jan25 would report Y and Z, which would not be known to HUMAnN 4, and so you would not get any X, Y, or Z features in your HUMAnN 4 output (X missing because MetaPhlAn 4 on Jan25 doesn’t report it, Y and Z missing because HUMAnN 4 on Oct 22 doesn’t have pangenomes for them). That would affect both the stratifications and the totals.
Hence I would recommend sticking with MetaPhlAn 4 on Oct22 for HUMAnN 4 until HUMAnN 4 is properly updated to Jan25.
I see, that makes sense. I will sticks to your recommendation. Thanks again for your help!