DESeq Output Plot and Options

[silverlining]

I analyzed my data using DESeq - 4 samples - in duplicates - for a total of 8. While I was able to pull out several DE genes, the plots look odd - attached. Please let me know your thoughts.

Also, initially, I did not have replicates and ran the analysis on just 4 samples. Oddly, my analysis with the replicates - for one pair - pulled out a lot more DE genes but primarily downregulated. I would not expect there to be such a bias towards downregulated genes only with the technical replicates - does this strike anyone as abnormal?

Attached is a venn diagram depicting how my initial analysis without replicates compared with my second analysis with replicates - oddly, there is very little overlap between the genes upregulated from the initial analysis (no rep) compared to with rep. Please give me your input as this is quite disturbing to me!

Finally, where can I find options to play around with the DESeq plot - ie. add a title, change colors etc.

I am a novice in this field so your input would be much appreciated.

Thanks!

[Simon Anders]

- You shouldn't give too much on the analysis wthout replicates. There is a reason why one shouldn't do such analyses, so don't be surprised if the proper analysis does not match the preliminary results.

- Well, you data does look asymmetric. Maybe your treatment causes predominantly downregulation.

- By the way, make sure that you deal correctly with variance outliers. See this thread: http://seqanswers.com/forums/showthread.php?t=13165

- Changing the plot appearance: The plot commands used in the vignette are deliberately kept simply. You can sutomize them as you would for any R plot. To read up on the latter, see e.g. Paul Murrell's book.

[silverlining]

Thanks for the reply! What do you say about the MA plot on the run with replicates looking so non-uniform? Is this indicative of some error?

I appreciate your input!

[Simon Anders]

I'm not quite clear on what is what in your plots. In any case, if you are worried that there are several very strongly upregulated genes but no similar strongly downregulated genes, pick a few of these genes (R's 'identify' function is useful for this), look at their normalized count values to check whether the effect is consistent across replicates and that it is not a variance outlier. Then, consider whether the result makes biological sense. After all, there is no a priori reason why all experiments should give the same amount of up and down regulation.