Dear All, As I continue my exploration of RNASeq analysis options, it seems that some approaches are better than others depending on your data-set and goals. I was wondering if anyone can give me a feeling on when is better to use Voom vs EdgeR (or DESeq) and whether modelling the distribution is really that important vs just modelling the variance. My data set is very noisy, with very low signal and very few differentially expressed genes, so sensitivity is key to me. I have 4-9 biological replicates per condition and good depth of coverage I also have microarrays and extensive qPCR data for the exact same RNA samples that were sequenced. Thanks in advance for the suggesitons -- Lucia Peixoto PhD Postdoctoral Research Fellow Laboratory of Dr. Ted Abel Department of Biology School of Arts and Sciences University of Pennsylvania "Think boldly, don't be afraid of making mistakes, don't miss small details, keep your eyes open, and be modest in everything except your aims." Albert Szent-Gyorgyi [[alternative HTML version deleted]]

Hi, On Wed, Mar 20, 2013 at 2:30 PM, Lucia Peixoto <luciap at="" iscb.org=""> wrote: > Dear All, > As I continue my exploration of RNASeq analysis options, it seems that some > approaches are better than others depending on your data-set and goals. > I was wondering if anyone can give me a feeling on when is better to use > Voom vs EdgeR (or DESeq) and whether modelling the distribution is really > that important vs just modelling the variance. If you search the mailing list, you'll see Gordon talk about this a few times: http://search.gmane.org/search.php?group=gmane.science.biology.informa tics.conductor&query=smyth+edger+voom There's a lot of good stuff there that's worth reading. I'm almost reluctant to point you directly to the following email, as I hope you take the time to read up on different topics discussed around voom and edgeR, but this might answer your question a bit more directly: http://thread.gmane.org/gmane.science.biology.informatics.conductor/45 469/focus=45500 Where Gordon says: """ My feeling the moment is that edgeR is superior for small counts and but that voom is safer and more reliable for noisy heterogeneous data. Only edgeR can estimate the biological coefficient of variation (as we defined this in our 2012 paper). But we are actively working on both methods, and are open to what we find. """ > My data set is very noisy, with very low signal and very few differentially > expressed genes, so sensitivity is key to me. I have 4-9 biological > replicates per condition and good depth of coverage > I also have microarrays and extensive qPCR data for the exact same RNA > samples that were sequenced. Given that you have a gold standard in your hands, in the form of qPCR data, why not try several methods and see which one recapitulates the qPCR data most closely? I'm sure some people would be interested in reading a summary of your results if you do so ... and perhaps http://rpubs.com might be a good place to post it, if you don't have webspace of your own :-) -steve -- Steve Lianoglou Defender of The Thesis | Memorial Sloan-Kettering Cancer Center | Weill Medical College of Cornell University Contact Info: http://cbio.mskcc.org/~lianos/contact