edgeR issue:RNA seq

[niel]

I am facing an error at the step Trended Dispersion, although no error during Common dispersion estimation.

rownames(design) <- colnames (d)
> z<- estimateGLMCommonDisp(d , design)
v<- estimateGLMTrendedDisp(d , design)
Error in t(y) + prior.count.scaled : non-conformable arrays

Please suggest what could be the reason, and what would be the way out.

[niel]

do I have to edit aveLogCPM.default, if yes, then How?

[bruce01]

I think

v<- estimateGLMTrendedDisp(d , design)

should be

v<- estimateGLMTrendedDisp(z , design)

ie you have to run CommonDisp on 'd', and then run TrendedDisp on the output of that.

[niel]

Ok.. yah.. Thanks. I was making some silly input mistake in some earlier step too. That step, including your suggession make it work properly. But strangely, all 20 top tags are showing Downregulation in disease. That has made me a bit sceptical. I am doing double check now.

Thanks

[bruce01]

Good that it works; to check your top 20 tags I suggest you look at FPKM/RPKM or raw counts to make sure the 'direction' of regulation is correct. EdgeR gives "-logFC" for the first condition you provide IIRC. This can be confusing, and I still double check all results against FPKMs.