[BioC] affyPLM interpretation

[Arne.Muller at aventis.com]

Hi All,

I've some question regarding the affyPLM package. Maybe you can give some
hints ... . 

fitPLM is *not* a normalisation method, is it? I mean, it performs a
normalisation such as RMA background correction and e.g. quantile cross-chip
normalisation, but it doesn't summarize the probes that belong to one probe
set into a single expression value. Instead, a robust (?) linear model is fit
through *all* the probes of each probeset on each chip. Is this kind of
interpretation correct?

How do I then interpret the coefficients of the PlmSet object? Say I've a 40
chips for measuring gene expression after 4h and 24h treatment with a drug
with doses 0mM, 0.1mM, 0.25mM, 0.5mM and 1.0mM (this is a typical design for
me).

I'd create factors 

time <- factor(c('04h','24h'))
dose <- factor(c('0mM', '0.1mM', '0.25mM', '0.5mM', '1.0mM'))

and then do the fit with an intercept

plm <- fitPLM(affybatch, model = PM ~ probes + dose + time)

I'm not sure what to do with the coefficients of the result, what do they
tell me?

A while ago I've analysed my data with linear model and anova in a way like

foreach gene in genes:
   mylm <- lm(intensity ~ dose + time, data=all_chips_dataframe_for_gene)
   myanova <- anova(lm)

Then I've extracted the p-values for the dose and time factor for each gene
to see what's differentially regulated ... .

In the example above I'm using the linear model for an anova - which makes
sense to me, and again the coefficients of "lm" wouldn' tell me much. Could I
use the the PlmSet for an anova, too?


	thanks for your comments
	+kind regards,

	Arne