Hi everyone,
I am performing an allele-specific RNA-seq analysis. Basically, I used SNP positions between two alleles of a mouse hybrid strain to separate reads specific to each allele. Now I normalized the data, as usual, using Deseq2 function and performed differential expression analysis.
However, I wonder if the normalization method for allele-specific counts will be the same or different from regular RNA-seq. If I consider only sequencing depth normalization (CPM: counts per million), the regular RNA-seq data will be normalized by total mapped reads and multiplied by 1M.
But now since I already separated my reads specific to two different alleles, does the same normalization method will make sense? In this case, instead of the total mapped reads, I used total reads assigned to each allele and normalized following counts per million (CPM) method.
I read a few things in this post about a similar topic but still not clear to me how should I proceed?
Please suggest.
Thanks
Be careful about nomenclature. CPM does not necessarily mean that libraries are solely scaled by library size.
edgeR
uses the term for its default normalization in which the library size is further corrected by a correction factor that accounts for library composition changes. The method is called TMM and implemented e.g. inedgeR::cpm()
I don't think that's right. I've got files made with EdgeR's cpm, and the output is the # counts / # of total assigned counts. No fancier math than that. The documentation for EdgeR's cpm doesn't mention TMM either. If you make a sample, as in the documentation for the function, it's just the simple division.
By default library size is the
colSum
, you are right on that one. Still, if the normalization factors are present in theDGElist
object they will be used to multiply thecolSums
with to correct for composition changes between samples.This is the
cpm
source code from lines 4 to 12:The same would hold true for
DESeq2::fp(k)m
, where you can choose between colSums or size factors for correction.Hi,
If anyone has a suggestion please let me know.
I would appreciate any help.
Thanks
Hi,
I would appreciate any suggestion.
Thanks