# [BioC] Creating a model matrix for three age groups using two-colour arrays

Gordon K Smyth smyth at wehi.EDU.AU
Thu Nov 20 23:00:05 CET 2008

```Dear Carthika,

This is a somewhat special design, and so isn't specifically covered by
the limma User's Guide.  It's like a direct design for LTP, but like a
common reference design for comparing the grous.

I suggest

agegroup <- factor(targets\$Name)
design <- model.matrix(~0+agegroup)
colnames(design) <- levels(agegroup)

Then the three coefficients compare LTP to Control within each of the
three age groups.  To compare the age groups, just extract contrasts as
usual,

cont.matrix <- makeContrasts(YvsA=young-aged,levels=design)

and so on

Best wishes
Gordon

> Date: Thu, 20 Nov 2008 12:13:13 +1300
> From: Carthika Luxmanan <carthika.luxmanan at anatomy.otago.ac.nz>
> Subject: [BioC] Creating a model matrix for three age groups using
> 	two-colour arrays
> To: bioconductor at stat.math.ethz.ch
> Content-Type: text/plain
>
> Hi
>
> I am new to Bioconductor R/Limma. I've worked through some examples,
> and familiarized myself with some basic commands. However, I am not
> sure how to create a model matrix for my experiment. We looked at
> three age groups of rats (YA, MA and OA) to look at gene expression
> changes following LTP (LTP is a memory model). LTP was stimulated on
> one side of the brain, while the opposite side served as control. So
> RNA from Control side was coupled to Cy3, while RNA from LTP side was
> coupled to Cy5. A single microarray was used for each animal, with n=5
> for YA, 3 for MA and 5 for OA. My Targets file looks as follows:
>
> SlideNumber	Name	FileName	Cy3	Cy5
> 1	young	YA19_with norm_15Sept06.gpr	Ct	LTP
> 2	young	YA23_with genepix norm.gpr	Ct	LTP
> 3	young	YA 24_with norm.gpr	Ct	LTP
> 4	young	YA27_with norm_15Sept06.gpr	Ct	LTP
> 5	young	YA28_with norm_15Sept06.gpr	Ct	LTP
> 6	middleaged	MA3_with norm_15Sept06.gpr	Ct	LTP
> 7	middleaged	MA12_with norm_15Sept06.gpr	Ct	LTP
> 8	middleaged	MA15_with norm_15Sept06.gpr	Ct	LTP
> 9	aged	OA14_with norm.gpr	Ct	LTP
> 10	aged	OA 19_with norm.gpr	Ct	LTP
> 11	aged	OA 32_with genepix norm.gpr	Ct	LTP
> 12	aged	OA37_with norm_150906.gpr	Ct	LTP
> 13	aged	OA38_bal_15Sept06.gpr	Ct	LTP
>
> How can I differentiate the groups in a model matrix? I would like to
> compare the gene expression changes between the three groups. I would
> have liked to look at the gene expression changes as a result of LTP
> in the individual groups too, but I have been warned recently that
> this might bring up more false positives since I do not have dye swaps
> within each group.
>