Cuffdiff significance

differential · 03-01-2013, 07:55 PM

Hi,

I have 3 sequenced transgenic Arabidopsis genomes (sequenced sing Tuxedo suite). I am looking for genes that exhibit significant differential expression in all 3 transgenic lines and ideally do not show significant differential expression in our WT control lines. When I use only genes that have an FDR<0.05 I get virtually no interesting genes (i.e., unknown proteins), but when I arbitrarily choose a fold change of 2 (chosen by my professor) and ignore the FDR, I get quite a few genes that are interesting to my professor.

Is there a way to justify using these genes even though cuffdiff says that the expression is insignificant?

Please let me know if this is not information, etc. I am very new at this.

rboettcher · 03-04-2013, 04:54 AM

Cuffdiff is known to be very stringent, especially if the variance within your groups is quite high. If you are accustomed to R, then you can try to use methods such as DESeq or edgeR for differential expression analysis, otherwise, just disregard Cuffdiff's FDR and set thresholds for fold change and overall expression and see how far you get (though I would not recommend that).
I also recently read about another RNAseq pipeline called Grape, maybe that helps you further. See http://bioinformatics.oxfordjournals.../29/5/614.long

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
SAM: Significance Analysis of Microarrays	Amative	Bioinformatics	0	11-27-2012 09:41 AM
Cuffdiff DE significance of zero FPKM values	jwalker_tgi	Bioinformatics	7	10-13-2012 04:14 PM
Significance of an FPKM value	JueFish	Bioinformatics	5	10-12-2012 12:27 AM
Determining clinical significance of variants	Rocketknight	Bioinformatics	2	09-28-2012 09:54 AM
read numbers AND biological significance	lwgbioinfor	Bioinformatics	0	07-18-2011 07:13 PM

03-01-2013, 07:55 PM	#1
differential Junior Member Location: Corvallis OR Join Date: Oct 2012 Posts: 1	Cuffdiff significance Hi, I have 3 sequenced transgenic Arabidopsis genomes (sequenced sing Tuxedo suite). I am looking for genes that exhibit significant differential expression in all 3 transgenic lines and ideally do not show significant differential expression in our WT control lines. When I use only genes that have an FDR<0.05 I get virtually no interesting genes (i.e., unknown proteins), but when I arbitrarily choose a fold change of 2 (chosen by my professor) and ignore the FDR, I get quite a few genes that are interesting to my professor. Is there a way to justify using these genes even though cuffdiff says that the expression is insignificant? Please let me know if this is not information, etc. I am very new at this.

03-04-2013, 04:54 AM	#2
rboettcher Member Location: Berlin Join Date: Oct 2010 Posts: 71	Cuffdiff is known to be very stringent, especially if the variance within your groups is quite high. If you are accustomed to R, then you can try to use methods such as DESeq or edgeR for differential expression analysis, otherwise, just disregard Cuffdiff's FDR and set thresholds for fold change and overall expression and see how far you get (though I would not recommend that). I also recently read about another RNAseq pipeline called Grape, maybe that helps you further. See http://bioinformatics.oxfordjournals.../29/5/614.long