Low Community Composition

23 views
Skip to first unread message

adamber1187

unread,
Jan 15, 2020, 3:40:53 PM1/15/20
to HUMAnN Users
Hello all,

I'm running HumanN2 with shotgun metagenome samples (150 bp). On the Metaphlan2 step the amount of species for my first sample is 81, but only explains 30.78% of predicted community composition. The input is about 5.5 million base pairs. Just curious why the amount explained is so low? When I've done this in the past it's usually 99-100%.

Thanks in advance,
Adam

Eric Franzosa

unread,
Jan 15, 2020, 4:03:08 PM1/15/20
to humann...@googlegroups.com
My first guess would be that a lot of predicted abundance is assigned to "Genus_unclassified" levels. These don't have pangenomes in HUMAnN2 to be selected, hence the abundance of the selected pangenomes adds up to a lot less than 100%.

Thanks,
Eric



--
You received this message because you are subscribed to the Google Groups "HUMAnN Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to humann-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/humann-users/6a803b6b-22fe-46b7-95eb-af1cf1e7a2df%40googlegroups.com.

adamber1187

unread,
Jan 16, 2020, 4:01:46 PM1/16/20
to HUMAnN Users
Hi Eric,

Thanks for the reply. Is this considered a normal thing? In the past when I've used Humann2, the composition was close to 100%. All of my samples now are 15-30%.

-Adam


On Wednesday, January 15, 2020 at 2:03:08 PM UTC-7, Eric Franzosa wrote:
My first guess would be that a lot of predicted abundance is assigned to "Genus_unclassified" levels. These don't have pangenomes in HUMAnN2 to be selected, hence the abundance of the selected pangenomes adds up to a lot less than 100%.

Thanks,
Eric



On Wed, Jan 15, 2020 at 3:40 PM adamber1187 <adam.be...@gmail.com> wrote:
Hello all,

I'm running HumanN2 with shotgun metagenome samples (150 bp). On the Metaphlan2 step the amount of species for my first sample is 81, but only explains 30.78% of predicted community composition. The input is about 5.5 million base pairs. Just curious why the amount explained is so low? When I've done this in the past it's usually 99-100%.

Thanks in advance,
Adam

--
You received this message because you are subscribed to the Google Groups "HUMAnN Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to humann...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages