[BioC] Limma, lmFit "subscript out of bounds" error

[Gordon K Smyth]

> Date: Tue, 10 May 2005 18:57:51 +0100
> From: Emmanuel Levy <emmanuel.levy at gmail.com>
> Subject: [BioC] Limma, lmFit "subscript out of bounds" error
> To: bioconductor at stat.math.ethz.ch
>
> Hello,
>
> I am working with 5 affymetrix arrays: 1 wild type and 4 mutants, all
> differents (no replicates).
> I think that is the reason why when I run lmFit I get a "subscript out
> of bound" error.
> I guess one can't fit a linear model with 1 point only.

The code you give shouldn't give an error, and doesn't in limma 1.9.5, but there isn't any point
in fitting a linear model to data with no replication.  The fitted model will just return your
expression matrix back again as coefficients.

> If it is the case, is there any way around? What would you recommend
> to analyse these arrays?

You have 5 different conditions and one array for each condition.  The only thing you can do is to
compute fold changes (or log-ratios) between the different conditions, and choose an arbitrary
cut-off to select genes.  This seems intuitively obvious, doesn't it?

Gordon

> Thank you very much in advance for your help,
>
>    Emmanuel
>
> ---------- more details -------------------------------------------------------
> R: 2.0.1
>
> Package:       limma
> Version:       1.8.16
> Date:          2005/01/09
>
> Package:       affy
> Version:       1.5.8-1
> Date:          2004-10-20
>
>
> My phenoData is as follow:
>> pData(eset2)
>      pheno
> C1 WT
> C2 MU1
> C3 MU2
> C4 MU3
> C5 MU5
>
> The design matrix as follow:
>> design
>   WT MU1 MU2 MU3 MU4
> 1  1   0   0    0       0
> 2  0   1   0    0       0
> 3  0   0   1    0       0
> 4  0   0   0    1       0
> 5  0   0   0    0       1
> attr(,"assign")
> [1] 1 1 1 1 1
> attr(,"contrasts")
> attr(,"contrasts")$f
> [1] "contr.treatment"
>
> When I run lmFit, I get an error:
>> fit = lmFit(eset2, design)
> Error in inherits(x, "data.frame") : subscript out of bounds