Hello Folks,
I am analysing a dataset using edgeR and DESEq2 and trying to understand what I am seeing.
Here is what I have so far:
I do not have any technical replicates of my samples, unfortunately it is just how the experiment was designed. All I have is control and treated(2 types of treatment)condition per sample and thats it...I use all the controls (as replicates ,across 2 types of treatment) and all the treated as replicates(across both types of treatment) together.
So I plug the HTSeq counts for these in DESeq2 and EdgeR and try to see how things compare.
DESeq2: worked fine in the sense it convinces me enough to be a method of choice in prifiling differentially expressed genes and I have 24 of them and clustering indicated that there is no effect of treatment and biology of samples play a high role.
EdgeR: I get about 40 differentially expressed gene according to my results from Common dispersion estimate but the tagwise dispersion estimate does not work at all !
My prior.n is 3.85.so when I do :
cds<-estimateTagwiseDisp(cds,prior.n=3)
Error in estimateTagwiseDisp(cds,prior.n=3)
unused argument (prior.n=3)
I tried using the default prior.n as 10 and it still does not work.
So the mean variance plot indicated that the common dispersion (solid blueline) and the tagwise dispersion(blue circles) to be tightly knit and the raw varainces along the same line but a little more dispersed than the former two.
So my question is:
Why is then the tagwise dispersion not working ?
Can I use the MDS plot from edgeR and the clustering heatmap from DESeq2 on the same comparable page? they both cluster the datasets however am I comparing apples to oranges here?
Also EdgeR does this CPM and then also has the pvalues from the cmn.dispersion results , can I compare this to the pvalues I got from results from DESeq2?
Pvalues are pvalues, however I was owndering if they are indicating the same things or atleast close enough that I can compare my diff.expressed genes from these two results to say edgeR gave me a,b,c and DESeq2 gave me b,c E,F and so on?
If the p-values are comparable, at least the way they are generated.......is it because the cooks distance is accounted for that I have less number of DE genes in DESe2 than in edgeR?
I am not a statistical person, so please if any one can comment on these , I would appreciate it. Good to get different perspectives.
Thanks
geneart.
I am analysing a dataset using edgeR and DESEq2 and trying to understand what I am seeing.
Here is what I have so far:
I do not have any technical replicates of my samples, unfortunately it is just how the experiment was designed. All I have is control and treated(2 types of treatment)condition per sample and thats it...I use all the controls (as replicates ,across 2 types of treatment) and all the treated as replicates(across both types of treatment) together.
So I plug the HTSeq counts for these in DESeq2 and EdgeR and try to see how things compare.
DESeq2: worked fine in the sense it convinces me enough to be a method of choice in prifiling differentially expressed genes and I have 24 of them and clustering indicated that there is no effect of treatment and biology of samples play a high role.
EdgeR: I get about 40 differentially expressed gene according to my results from Common dispersion estimate but the tagwise dispersion estimate does not work at all !
My prior.n is 3.85.so when I do :
cds<-estimateTagwiseDisp(cds,prior.n=3)
Error in estimateTagwiseDisp(cds,prior.n=3)
unused argument (prior.n=3)
I tried using the default prior.n as 10 and it still does not work.
So the mean variance plot indicated that the common dispersion (solid blueline) and the tagwise dispersion(blue circles) to be tightly knit and the raw varainces along the same line but a little more dispersed than the former two.
So my question is:
Why is then the tagwise dispersion not working ?
Can I use the MDS plot from edgeR and the clustering heatmap from DESeq2 on the same comparable page? they both cluster the datasets however am I comparing apples to oranges here?
Also EdgeR does this CPM and then also has the pvalues from the cmn.dispersion results , can I compare this to the pvalues I got from results from DESeq2?
Pvalues are pvalues, however I was owndering if they are indicating the same things or atleast close enough that I can compare my diff.expressed genes from these two results to say edgeR gave me a,b,c and DESeq2 gave me b,c E,F and so on?
If the p-values are comparable, at least the way they are generated.......is it because the cooks distance is accounted for that I have less number of DE genes in DESe2 than in edgeR?
I am not a statistical person, so please if any one can comment on these , I would appreciate it. Good to get different perspectives.
Thanks
geneart.
Comment