Question: metagenomeSeq log transformation
0
gravatar for PJ
2.3 years ago by
PJ0
United States
PJ0 wrote:

Hi,

Normalized counts from the matrix table are transformed to log2 values by metagenomeseq, aren´t they? My question is about the zeros in the count matrix. How are they transformed to allow statistical analysis? (as log2 of zero gives an error). Thanks!

ADD COMMENTlink modified 2.3 years ago by Joseph Nathaniel Paulson270 • written 2.3 years ago by PJ0

In addition, how is it possible that features with negative log2 values in the transformed count matrix come up with positive average expression values in the statistical output? Thanks!

ADD REPLYlink written 2.3 years ago by PJ0
Answer: metagenomeSeq log transformation
0
gravatar for Joseph Nathaniel Paulson
2.3 years ago by
United States
Joseph Nathaniel Paulson270 wrote:

To extract raw, or normalized counts, the MRcounts function is used. If you just type MRcounts you'll see that we use a continuity-correction pseudocount of 1. This maps 0s --> 0s in log-space.

ADD COMMENTlink written 2.3 years ago by Joseph Nathaniel Paulson270

Thanks Joseph for the quick answer. What about the negative tansformed log values? Is the program taking the absolute values for statistics? 

ADD REPLYlink written 2.3 years ago by PJ0
That means you're not putting counts in. Log of a fraction will lead to a negative number. 

On Jan 21, 2017 5:26 PM, "PJ [bioc]" <noreply@bioconductor.org> wrote:
Activity on a post you are following on support.bioconductor.org

User PJ wrote C: metagenomeSeq log transformation:

Thanks Joseph for the quick answer. What about the negative tansformed log values? Is the program taking the absolute values for statistics? 


Post tags: metagenomeseq, logtransformation

You may reply via email or visit C: metagenomeSeq log transformation


ADD REPLYlink written 2.3 years ago by Joseph Nathaniel Paulson270

Yes, you are right. I added a second normalization step based on reference gene length and 16S bacterial reads so I am not using the read counts. Could this affect the statistical output? Where can I see the actual formula to calculate average expression? Thanks again!

ADD REPLYlink written 2.3 years ago by PJ0

This normalization step is not expected input for metagenomeSeq and it was not designed for that. The method was designed for read counts that you can read up on in the supplementary material here: http://www.nature.com/nmeth/journal/v10/n12/full/nmeth.2658.html. Or in the vignette.

ADD REPLYlink written 2.3 years ago by Joseph Nathaniel Paulson270
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 126 users visited in the last hour