RNA-Seq normalization for co-expression analysis
1
0
Entering edit mode
Lin ▴ 50
@lin-19103
Last seen 4.2 years ago

Hi all,

it is my first time to work with RNA-seq data, and with this data a differential expression analysis and co-expression network analysis should be done. Now I read in the pipeline [RNA-seq analysis is easy as 1-2-3 with limma, Glimma and edgeR] for differential expression analysis that at first the counts are filtered (CPM), and then normlization is done with TMM. However, for the co-expression analysis I would like to use normalized data but another filtering method. So my question is: Can I also apply the TMM normalization method to unfiltered data, and then filter the normalized data afterwards? Or do you see any problem with this/have other suggestions?

Thanks in advance!

edger WGCNA limma • 2.2k views
ADD COMMENT
0
Entering edit mode
ADD REPLY
0
Entering edit mode

Hi Kevin, thanks for your answer! But my question was rather if there is a problem using TMM normalization FIRST (with unfiltered data), and filter the data afterwards (because I want to use another pipeline where the whole filtering is implemented).

ADD REPLY
1
Entering edit mode

I see. Gordon has already answered. Based on your logic, you have CPM counts, and then you apply TMM to those?

ADD REPLY
0
Entering edit mode

Yes, exactly, that would be what I thought of... Because I need normalized data, but would like to use another filtering procedure during the co-expression analysis... And with this filter I would loose too many transcripts before (and would double-filter).

ADD REPLY
0
Entering edit mode

Typically, we filter the raw counts, then normalise, and then make statistical inferences on the normalised counts. After that, we may apply a further transformation on the normalised counts for the purposes of conducting downstream analyses.

ADD REPLY
1
Entering edit mode
@gordon-smyth
Last seen 36 minutes ago
WEHI, Melbourne, Australia

Yes, you can apply TMM to the unfiltered counts, although it is not quite a robust as applying it to the filtered counts.

But filtering out genes that are not expressed to a meaningful degree in any sample would still be sensible as a first step, as this is needed by both co-expression analyses and edgeR.

ADD COMMENT

Login before adding your answer.

Traffic: 548 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6