Entering edit mode
Dear Agnes,
> Date: Wed, 08 Jan 2014 11:45:41 +0100
> From: Agnes Paquet <paquet at="" ipmc.cnrs.fr="">
> To: bioconductor at r-project.org
> Subject: [BioC] Question about best analysis method for a complex
> array expriment design
>
> Dear List,
>
> I need to analyze an experiment with a design more complex than
usual
> for our facility, and I am not sure about the best way to analyze
this
> dataset. I would really appreciate your advice on whether the method
I
> am planning to use is correct, or if there is a better way to
analyze
> this data.
>
> The experiment design is the following:
> We have 10 patients, and we are using one-color Agilent arrays. Each
> patient performed a physical test twice: once without anything
added,
> and once taking a drug during the test. Samples are collected before
and
> after the physical test, for a total of 4 samples by patients. The
drug
> was administered randomly during the first or second test.
>
> Here is the top of my target file:
>
> Patient.ID TimePoint Drug TestOrder
Drug.Included.In.Test
> Pt1 Before no test1 control
> Pt1 After no test1 control
> Pt1 Before no test2 test
> Pt1 After yes test2 test
> Pt2 Before no test2 control
> Pt2 After no test2 control
> Pt2 Before no test1 test
> Pt2 After yes test1 test
>
> We are interested in finding:
> - DE genes related to physical test only
> - DE genes related to the addition of the drug only
> - Genes differentially regulated by the drug during the physical
test
>
> I usually use limma for differential analysis, so following the
limma
> user?s guide, I was planning to use a design with blocks of size 4
for
> patients, and a variable with 4 levels combining
Drug.Included.In.Test
> and TimePoint.
>
> Is this approach correct?
Yes.
> I read in the user?s guide patient information could also be modeled
as
> random effect using the duplicateCorrelation function. Would this
method
> be more appropriate?
Would not give an advantage here (balanced design and only two
patients).
> Is there a better way to model the data, that would estimate the
> physical test effect and the drug effect directly?
No.
Best wishes
Gordon
> Thank you very much for your help,
>
> Agnes
______________________________________________________________________
The information in this email is confidential and
intend...{{dropped:4}}