[BioC] normalization and clustering etc...

[Phguardiol at aol.com]

Hi all,
I have recently discover Cyber-T and I was wondering if this Bayesian 
approach for comparison was available in Bioconductor since I dont always 
have a fast internet connection and I m not working on Unix / Linux yet ? 
What seems to be a similar approach is apparently available in the new S+ 
module Array analyzer is it also in R/Bioconductor ? 
In the same way, is there any kind of Bayesian clustering method available in 
Bioconductor like the one available in Array Miner ?
   
In addition, I have two questions that I d like to have some help about from 
your group:
I ve affy 133A+B chips made from RNA coming from 4 cell lines. Two of these 
cell lines are B cells (one has a deficient gene and the other has the 
corrected gene inserted instead of the deficient one, both are coming from 
the same parental cell line) and 2 others are fibroblastic cell lines in 
which the same gene is inactivated or not. Therefore I have 2 cell lines 
deficient for a gene C (one fibroblast one B cell) and their corrected 
homologs. In addition, all these cells have been exposed to the same 
chemotherapeutic agent at the same dose same duration so that I have data 
from untreated and treated cells. I am planning to compare corrected versus 
uncorrected untreated B cells, then the same for fibroblasts, then all this 
again for treated cells. The aim is to isolate genes that are differentially 
expressed in nornal conditions between the deficient and corrected B cells 
then the same for the fibroblasts and then finally to see which one are in 
both cases differentially expressed. Finally I d like to know what are the 
modifications of these results when I stress the cells with chermo. So here 
are my qurestions:
- For normalization: should I normalize the whole set of chips then do all my 
comparisons or should I normalize only the chips I am planning to compare at 
each time and repeat this process for each comparison ? I m planning to use 
the RMA module.
- For clustering: I have identified at least one gene W of great interest in 
my model and I was planning to do clustering analysis to see what are the 
genes with a similar pattern of expression. Using all my differents 
conditions with my 4 cell lines I found one very interesting gene Z in the 
cluster of the first gene W. I was wondering if it would be a good idea to 
add some extra chips made from completely different cells, ie, CD34+ 
hematopoietic stem cells, mature Cd22+ B cells, acute leukemia cell lines in 
my case, as a validation process ? For instance, if the gene Z is not anymore 
close to gene W with these extra chips... what could be the conclusion...? 
Thanks for your help
Philippe

	[[alternate HTML version deleted]]