Question: DESeq2: Normalization of allele specific counts in RNA-seq
gravatar for Ankit
19 days ago by
Ankit0 wrote:

Hi everyone,

I am posting here about the normalization of allele-specific counts in RNA-seq . I have previously asked this question at another platform. And finally reached here after the suggestion. I have read this manual by Deseq2 author. This is very well written manual on how to use Deseq2 for allele-specific analysis. I have some queries on how I can better use Deseq2 for my allele-specific analysis.

What I understood that author did not use size factor for normalization rather set the size factor to unity and calculated alternate vs reference allele ratios.

  1. If we set the size factor to unity, the counts will remain same. Then how to extract normalized count matrix?

  2. Using alt / ref ratio we can identify differentially expressed genes in allele -specific manner between control and treated samples. In this case, how to find differential expression between the possible combinations:

control-allele1 vs control-allele2

control-allele1 vs treated-allele1

control-allele2 vs treated-allele2

treated-allele1 vs treated-allele2

In the manual resultsNames(dds) information are as follows:

[1] "Intercept" "conditiontreatedvs_control" [3] "conditioncontrol.sample2" "conditiontreated.sample2" [5] "conditioncontrol.sample3" "conditiontreated.sample3" [7] "conditioncontrol.sample4" "conditiontreated.sample4" [9] "conditioncontrol.sample5" "conditiontreated.sample5" [11] "conditioncontrol.countalt" "conditiontreated.countalt"

conditioncontrol.countalt gives the alt / ref ratio in control and conditiontreated.countalt will gives the alt vs ref ratio in treated samples. So, might be DE genes can be obtained for combinations,

conditioncontrol.countalt gives the alt / ref ratio = control-allele1 vs control-allele2

conditiontreated.countalt gives the alt / ref ratio = treated-allele1 vs treated-allele2

There is one command to call out interaction between control and treated sample:

res.diff <- results(dds, contrast=list("conditiontreated.countalt", "conditioncontrol.countalt"))

But I don't how to extract the similar results for

control-allele1 vs treated-allele1

control-allele2 vs treated-allele2

Please suggest. I would appreciate any help.


ADD COMMENTlink modified 17 days ago by Michael Love25k • written 19 days ago by Ankit0

Hi everyone,

Any suggestion would be helpful.


ADD REPLYlink written 18 days ago by Ankit0
Answer: DESeq2: Normalization of allele specific counts in RNA-seq
gravatar for Michael Love
17 days ago by
Michael Love25k
United States
Michael Love25k wrote:

1) DESeq2 won't have normalized counts in the case that you set sizeFactors to 1. The analysis is on the raw counts, and the alt vs ref for each sample.

2) Your code looks correct for the within condition comparisons and the across condition comparison.

You can't compare allele 1 across control and treatment, that's not supported by this design, because of how we control for the patient baseline differences. If you want to understand this in more detail, I'd recommend consulting with a statistician or bioinformatician who has experience with linear models.

ADD COMMENTlink written 17 days ago by Michael Love25k

Thank you sir for the reply.

I agree with your valuable suggestions. I was thinking an alternative approach to get normalized count matrix. May be you can correct me on this. So I have two types of matrix. One with allele-specific counts (eg. Sample1 - allele1, allele2) and other before splitting (eg. Sample1 only). I thought to calculate size factors from the matrix before splitting. This size factor I want to set for both allele1 and allele2 for same sample(eg. Sample1). Then if I run Deseq(), it will normalize using this size factor and I can extract the normalised counts.

Or other way instead of using count matrix from before splitting, I sum up two allelic counts from matrix after splitting and then calculate size factors and use it downstream. Do u think it will make sense?

Please suggest.


ADD REPLYlink modified 17 days ago • written 17 days ago by Ankit0

Hi Sir, If you have any suggestions please let me know. I don't know how to proceed. I thought for some possible approaches to get normalized count matrix as I mentioned in reply. I am not sure if these approaches are appropriate. Please help. Thank you

ADD REPLYlink written 13 days ago by Ankit0

I’d recommend working with a statistical collaborator to design your analysis plan. I don’t have any more specific recommendations.

ADD REPLYlink written 12 days ago by Michael Love25k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 275 users visited in the last hour