I am trying to do an analysis for differential gene expression for rna-seq data, cell line hela for different time points. For example this is 6 to 0 hours. People told me to do the analysis with no shrinkage but i dont believe its the right way to do it
for example a gene with shkrinkage is up regulated with fold change : 11,18 and without shrinkage : 9.429,02 .
I mean this isnt even biologically correct right ?
If someone knows please enlighten me wether to use shrinkage or no. And how biologically correct is seeing a fold chage of 9.000 even 1.000.
edit with q-pcr they found that this gene up regulated 1000 times
People told me to do the analysis with no shrinkage
This does not sound very scientific. Given that the DESeq2 package and the authors behind it most likely can be considered to be among the top experts when it comes to RNA-seq analysis, you should have a good reason to use other than the default DESeq2 settings for a standard analysis. Did the "people" give any explanation for their advice?
Also (as an advice), to avoid confusion, please use "." as float point separator and no commas for decimals.
Applying the shrinkage definitely changes the result, otherwise what would be the point of doing it in the first place. As Grant has pointed out, DESeq2 is mostly trying to address the problem that lowly expressed genes tend to have high relatively levels of variability and they try to alleviate this.
The goal of a garden-variety RNA-seq experiment is usually to discover genes that show significant changes. DESeq2 and the other tools have been tuned to do exactly that in a fairly reproducible and reliable manner. Yes, they also give an estimate of the (log2) fold change, but they make it very clear that this an estimate either way.
Even without that clear language that can be found in the respective manuals, I wouldn't bet a single penny on a fold change value stemming from an RNA-seq analysis to be taken at face value. There are so many factors that will influence it, including the library prep.
Trying to compare these values precisely to qPCR measurements is not appropriate either because qPCR typically measures one (!) very short fragment of a transcript whereas RNA-seq often covers the entire transcript. EDIT: and thousands different other ones, too!
In short, I don't think you need to worry about a difference of an extremely high fold change (if that's a log2FC, we're talking about more than thousand fold difference, which is definitely in the ballpark of the qPCR, no?). The important thing is that the general message between RNA-seq and qPCR is consistent, i.e. the gene should change in the same direction, and if the ranking fits (i.e., the gene changes quite strongly compared to many other genes), then that's even better.
The user manual of DESeq2 states "Shrinkage of effect size (LFC estimates) is useful for visualization and ranking of genes." I think it is particularly useful for lowly expressed genes because they have high variability and shrinkage can reduce it. And I am not sure how well it behaves for super differentially expressed genes.
Regarding the second part of your question, than I have observed very high fold changes (several thousand, which is experimentally validated) in fungi (don't know about human). But If the result is validated with q-pcr then it should be reliable.
EDIT: also pay attention that DESeq2 report the fold changes in log2.
This does not sound very scientific. Given that the DESeq2 package and the authors behind it most likely can be considered to be among the top experts when it comes to RNA-seq analysis, you should have a good reason to use other than the default DESeq2 settings for a standard analysis. Did the "people" give any explanation for their advice?
Also (as an advice), to avoid confusion, please use "." as float point separator and no commas for decimals.
"People" gave an explanation regarding the q-pcr information i gave above.
Whether it is scientific or not it doesnt matter. Thanks for the info though.