Hello everybody,

finally i want to compare the effect on gene expression of my original protein with a truncated version of the protein.
First i analysed the effect of the original protein.
After expression of the original protein for 6h, about 2000 genes are regulated according to Deseq (not DeSeq2) within Galaxy. I compared triplicates of cells, where the protein was not induced against triplicates of cells, where the protein was expressed for 6h.
After i did my experiment the very same way with the truncated protein, there was the new version of DeSeq, DeSeq2.
When i run the programm, the output told me, that only 4 genes are regulated, 1 of them is the control gene, which is an enormous outlier. Therefore the output of DeSeq2 was that the log2FoldChange is 1,66 for the outlier, even though it should be around 8.5. Therefore i was told to change the following parameter: betaPrior = FALSE. Now the output is fine.
I'm sure i have to use DeSeq2 for the data of the original proteine to make it compareable.
Nevertheless is was wondering if i have to analyse the data of the original protein also with the betaPrior = FALSE setting it keep it compareable? Or can i somehow spoil the result this way?

Thanks a lot for a advise, Alex

The code i got from the DeSeq2 specialist for R:

sampleTable <- data.frame(name=1:6, file=paste0(c(50,70,80,56,76,86),".tabular"), condition=factor(rep(1:2,each=3)))
dds <- DESeqDataSetFromHTSeqCount(sampleTable, design=~condition, directory=".")
dds <- DESeq(dds, betaPrior=F)
res <- results(dds)

My export:

png(file = "Original_protein_betaPriorF.png")
plotMA(res)
title("Gene_expression_original_protein_0_vs_6h_betaPriorF")
dev.off()

The following grafics show the original protein analysed without and with the betaPrior option,
as well as the truncated protein with and without the betaPrior option.

Original_protein_betaPriorF:

Original_protein_betaPriorT:

Truncated_protein_betaPriorF:

Truncated_protein_betaPriorT:

hi,

Because you have the single gene (positive control) which you know will strongly deviate from the rest of the genes in both analyses, the assumption that the LFC have a common normal distribution prior is not valid, and so you should use betaPrior=FALSE for both analyses. Note that the LFC which are larger in absolute value than 10, with low mean expression are genes with all zeros in one condition, so these values are headed to +/- infinity.

hi,

Because you have the single gene (positive control) which you know will strongly deviate from the rest of the genes in both analyses, the assumption that the LFC have a common normal distribution prior is not valid, and so you should use betaPrior=FALSE for both analyses. Note that the LFC which are larger in absolute value than 10, with low mean expression are genes with all zeros in one condition, so these values are headed to +/- infinity.