I run DEXSeq on my data. I have 16 individuals in each conditions (2 conditions). After computing the dispersions with estimateDispersions( dxd ), I plotted the graph plotDispEsts( dxd ) and I obtain this :

I tried the others ways to fit (local and mean) but I obtained the same kind of shrinkage.

Can you please tell me if my graph seems to be ok ? And if it's not, what it could be ?

Moreover, is there any way to recover the normalized data ?

Thanks for you help,

Have a look at the raw data (counts) of the exons with coverages 1...100 and very small dispersions. Where do they come from? This is unlikely to be genuine biological or unavoidable technological variation; it could be a bioinformatics data processing artifact that's fixable and better be. How did you produce the count table.