[BioC] (1) Anova for 12 Affy yeast chips (Yg98) // (2) Problem fitting Li&Wong model with Affy 1.0 :

julien.sylvestre@wotan.ens.fr julien.sylvestre@wotan.ens.fr
Sun, 05 May 2002 15:22:32 +0200


(1) Anova for 12 Affy yeast chips (Yg98)

I'm going to try to explain the problem in a better way. I have the following experimental design :

Fraction 1
Fraction 2  		Medium A
Fraction 3 	*	Medium B		*	2 Reps

- Fractions 1, 2 and 3 are three distinct subcellular mRNA populations ;
- Medium A and B are two distinct culture medium ;
- By Rep is meant a 'biological repetition'. That is cells harvested independantly from which the 3 Fractions are extracted. I have  two sets of three Fractions in Medium A and  two sets of three Fractions in
Medium B.
I am interested in finding - in a given medium - the distribution of each mRNA among the 3 Fractions.

I thougth the data quite suited for ANOVA, at least as a first analytic step.
The model would explain the log expression value by the sum of  :
- a constant term (maybe null actually)
- a gene effect : approximately 6000 levels
- a fraction effect : 3 levels
- a medium effect : 2 levels
- a rep effect : 2 levels for each medium
- and possibly all two-order interactions.

As wrote before, I couldn't apply directly 'lm' or 'aov' in R due to memory problems (the matrix to inverse being too large). Nor  could I use Matlab Stat toolbox 'anovan' function which resulted in a disaster or
the 'MANOVA' free matlab toolbox which is clearly specialised in cDNA array data. I am to try SAS 'proc anova' soon which should at  least provide me average effects (that  I admit are not so interesting by
themselves). Before I calculate the whole thing myself, as a last solution, I wanted to make sure nobody had the idea of a  particular tool to use directly or adapt easily. I looked rapidly at the 'Anova' function
in the Genefilter package but it seems to use lm and to be limited to a one-way analysis hence I didn't find how to take advantage  of it.

(2) Problem fitting Li&Wong model with Affy 1.0 :

I get almost the same problem wusing the cel.container mechanism (actually I didn't try it before because I thought I was to use a  command called read.cel.container which in fact does not seem to exist) :
> myfiles = c("xxx1.cel", "xxx2.cel",...)
> lcel = read.container.celfile(filenames = myfiles
> cdf = read.cdffile("Yg_s98.CDF")
> eset = generateExprSet(lcel, cdf, method = "liwong", normalize = F)
> eset = generateExprSet(lcel, cdf, method = "liwong")

	9335 ids to go process...
	instancianting an exprSet...
	Error in data.matrix[, !phi.outliers] %*% phi[!phi.outliers] :
       		 non-conformable arguments

> traceback()
	4: which(cdf@name == i, arr.ind = TRUE)
	3: locate.name(ids[id], cdf, type = "pm")
	2: .local(x, cdf, method, ...)
	1: generateExprSet(lcel, cdf, method = "liwong")

Thank you.

JS.