How is the relab in humann2_renorm_table calculated

[Angel Dai]

Hi,

I have a question about how is the relab in humann2_renorm_table calculated. 

Since I have seen in one paper they are using the "cpm" as the relative abundance.

And also for input to the lefse, should I use the relab or the cpm normalized values?

Thank you very much for your time and attention!

Regards

Angel

[Eric Franzosa]

In the HUMAnN2 context, CPM refers to "copies per million" and is just relative abundance multiplied by 1 million to get "better-looking" numbers. In the sense that better-looking numbers can be more numerically stable, I would use the CPMs in LEfSe over relative abundance.

Also note that HUMAnN2's CPMs account for sequence length, as they are normalized starting from RPK units. CPM is sometimes used to refer to "counts per million", which does not imply that length-normalization has occurred.

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/8d1e454d-7dc7-48e2-9f0b-d99352ab8646%40googlegroups.com.

[Dai, Anqi]

Hi Eric,

How are you doing?

Just wanna confirm that, the CPM is calculated as the count in RPK that is output by humann2 divided by the library size of that sample and then times a million right?

So it doesn't matter which samples are included when you are not doing the renorm?

Or if I'm removing some samples after renorm, I don't need to do the renorm again starting from a cohort with my updated samples?

Thank you very much!

Angel 

---

From: humann...@googlegroups.com <humann...@googlegroups.com> on behalf of Eric Franzosa <fran...@hsph.harvard.edu>
Sent: Wednesday, October 30, 2019 2:54 PM
To: humann...@googlegroups.com <humann...@googlegroups.com>
Subject: Re: How is the relab in humann2_renorm_table calculated

 

To view this discussion on the web visit https://groups.google.com/d/msgid/humann-users/CAC%3DWeKore%3D8d6Wao_kZuSdM2WUFgQ5KmwF2xoNRCpvbc0p%2BUmA%40mail.gmail.com.