Hello everyone,
When comparing gene expression levels between two tissue samples, differences may arose for transcripts that pertain to cell subpopulations which are not in the same proportion in each of the compared samples. For instance, I am harvesting mouse olfactory epithelia. I will always get a diffrent proportion of cartilaginous tissue, mature and immature olfactory sensory neurons (mOSNs). Thus, when analysing RNASeq, differentail expresison analyses of mOSN transcript levels are affected by the quality of my dissection, and the biological differences between the mice. In qPCR, it is common to use normalization genes, ie. genes which expression pertain to the same cell type of the genes of interest, but which are not affected by the experimental condition. Did you already face the same problem? How would you normalize transcripts levels prior to differential expression analysis by DESeq or sleuth for instance?
Thank you in advance for any suggestion,
Joël
I just found out how to use normalization gene in DESeq:
The function DEseq runs the 3 following functions : dds <- estimateSizeFactors(dds) dds <- estimateDispersions(dds) dds <- nbinomWaldTest(dds)
(where dds is a DESeqDataSet object)
estimateSizeFactors is by default run on the whole dataset, but you can define control genes instead like estimateSizeFactors(dds, controlGenes=1:10) where (in this case) the 10 first genes in dds. So instead of using DEseq function, you run the tree functions with that modification of estimateSizeFactors and you get your dataset normalized according to a subset of genes.
Then, I guess that we should discard any gene that is not specific to the cell type of interest.