[BioC] Combining arrays arrays from different tissues and different labs.

[Kevin R. Coombes]

Hi,

Most normalization methods assume (either explicitly or implicitly) 
something about the distributions of values within each sample.  At the 
very least, they assume that some properties of these distributions are 
the same.  The case of mouse embryos and adult tissues illustrates one 
of the places these assumptions are likely to be violated.  For instance,
	[1] is the number of expressed genes the same?
	[2] are there roughly equal numbers of high, medium, and low expressing 
genes?
	[3] is the median expression level the same?
	[4] is the total RNA per cell the same?

Since genes are getting turned on and off at various rates in the 
development from embryo to adult, and since embryonic cells are actively 
proliferating and differentiating in ways that normal adult cells are 
not, it seems to me that there is no good reason to believe (without 
experimental data supporting it) that anything about the distributions 
are the same.

If you normalize within (but not between) the two data sets, you should 
be able to start to answer questions [1] and [2].  If the answers are 
"yes", then it makes sense to start thinking about how to normalize 
between the two data sets....

	Kevin

marco zucchelli wrote:
> Hi,
> 
> I would like to  check for differential expression on arrays
> downloaded from public databases.
> I was reading somewhere that it might be not very wise to normalize
> together arrays produced from different tissues and different labs.
> 
> In practice I would like to compare mouse embryos from one experiment
> with mouse adult tissues from another experiment.
> 
> Does somebody have an opinion on this ?
> 
> I tried to normalized the adult arrays and the embryo arrays
> separately, rescale them to the same median and then extract the
> expression values from all the arrays together (with li.wong method).
> I tried also to normalize and extract the expression values for all
> the arrays together with the same method. The results seem to be
> pretty similar.
> 
> In this case the amplification protocols used are different since
> embryo material is very small..
> 
> Best regards
> 
> 
> Marco
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor