Question

Random effects with limma's camera()

0

Entering edit mode

Stephen Hoang ▴ 30

@stephen-hoang-6156

Last seen 7.0 years ago

Hi All,

Is there a way to incorporate the consensus correlation calculated by duplicateCorrelation() into the gene set enrichment test performed by camera()? In short, I would like to treat one of my factors as a random effect via duplicateCorrelation() (this is what I did for differential expression testing). As far as I can tell, camera() is only aware of fixed effects. I don't have the option of treating my random effect as a fixed effect because I end up with a design matrix that is not of full rank. It seems like my only option is to run camera() using only fixed effects, and treat the results as overly-conservative.

Thanks in advance,

Steve

limma camera duplicatecorrelation • 2.2k views

ADD COMMENT • link updated 8.9 years ago by Gordon Smyth 52k • written 8.9 years ago by Stephen Hoang ▴ 30

score 1 · Accepted Answer · 2016-02-14

1

Entering edit mode

Gordon Smyth 52k

@gordon-smyth

Last seen 7 hours ago

WEHI, Melbourne, Australia

Sorry, you can't use duplicateCorrelation() with camera(). It has been on my to-do list for a while, but I haven't managed to find time to develop and implement it.

You can use duplicateCorrelation() with roast(), mroast() or romer().

ADD COMMENT • link 8.9 years ago Gordon Smyth 52k

2

Entering edit mode

Can this now be addressed by using the recent pre-ranked cameraPR() function? i.e. run lmFit() with block and correlation, then eBayes() (or treat() if appropriate?) and use the t-statistics.

ADD REPLY • link 7.2 years ago jamie.gearing ▴ 60

0

Entering edit mode

Yes, that's right (as long as you're happy with a preset value for the inter-gene correlation).