help! cannot find any deferentially expressed genes

[ericfit]

I have two groups, dose low vs dose high. # of dose high = 25, # of dose low = 51.

I used DESeq for the analysis. As the manual suggested, I have a quite large number of replicates, so for estimateDispersions, I chose sharingMode="gene-est-only".

Unfortunately, I got nothing, not even close to 0.05.

The variance plot looks like this:


As you may see, the variance is too large to get significant results. My question is what I can do next. Try few samples instead of too many replicates?

Any suggestion is welcomed.

[dpryan]

FYI, we can't see your variance plot.

A good rule of thumb in that case would to be to cluster your samples and see if there's any apparent batch effect that you need to account for. Alternatively but in a similar vein, have a look at the sva package. There's also always the possibility that there simply are no DE genes to be found.

[kopi-o]

dpryan's suggestions are good (look out for batch effects!), but in addition, as you have so many replicates, I would try SAMSeq (samr package in R) which usually works well in such a scenario.

[GenoMax]

FYI, we can't see your variance plot.

Here is the plot link Ericfit tried to post but was not correctly parsed: http://tinypic.com/r/2rc1jme/5

Ericfit: For future reference, when you are editing a message you can use the "Go advanced" mode. This will allow you to attach PNG images using the "paperclip" icon in the top menu bar.

[dpryan]

Here is the plot link Ericfit tried to post but was not correctly parsed: http://tinypic.com/r/2rc1jme/5

Ericfit: For future reference, when you are editing a message you can use the "Go advanced" mode. This will allow you to attach PNG images using the "paperclip" icon in the top menu bar.

Thanks GenoMax. Yikes, that's nasty variance!