[BioC] limma beginner needs help with matrix design

[Agnieszka Zmienko]

Hi!
I am looking for advice on how to get lists of 
differentially expressed genes for my 
experiments. Generally I have four plant lines 
(wt and 3 mutant lines) and two different 
conditions. But I do not have one common 
reference. I always compared treated plants to 
untreated pool of the same line and additonally I 
compared untrated mutants to wt pool. I performed 
4 replicates of each (biological replicates with 
respect to Cy5 channel). Background correction and normalization are done.

These are my arrays:

Cy3             Cy5
pool.wt         wt.cond1
pool.wt         wt.cond1
pool.wt         wt.cond1
pool.wt         wt.cond1

pool.wt         wt.cond2
pool.wt         wt.cond2
pool.wt         wt.cond2
pool.wt         wt.cond2

pool.mu1        mu1.cond1
pool.mu1        mu1.cond1
pool.mu1        mu1.cond1
pool.mu1        mu1.cond1

pool.mu1        mu1.cond2
pool.mu1        mu1.cond2
pool.mu1        mu1.cond2
pool.mu1        mu1.cond2

pool.mu2        mu2.cond1
pool.mu2        mu2.cond1
pool.mu2        mu2.cond1
pool.mu2        mu2.cond1

pool.mu2        mu2.cond2
pool.mu2        mu2.cond2
pool.mu2        mu2.cond2
pool.mu2        mu2.cond2

pool.mu3        mu3.cond1
pool.mu3        mu3.cond1
pool.mu3        mu3.cond1
pool.mu3        mu3.cond1

pool.mu3        mu3.cond2
pool.mu3        mu3.cond2
pool.mu3        mu3.cond2
pool.mu3        mu3.cond2

pool.wt         mu1.U
pool.wt         mu1.U
pool.wt         mu1.U
pool.wt         mu1.U

pool.wt         mu2.U
pool.wt         mu2.U
pool.wt         mu2.U
pool.wt         mu2.U

pool.wt         mu3.U
pool.wt         mu3.U
pool.wt         mu3.U
pool.wt         mu3.U

I am interested in:
performing all simple comparisons Cy5/Cy3
find difference in response to cond 1 and cond2 for each separate line
find differences in response to cond1 between all lines, the same for cond2
find differences between untreated mutant lines.

Can I do it by designing one smart matrix and 
making contrasts? or shall I better divide my 
tasks, for example into groups with one common 
reference each and perform the comparisons 
separately? But how do I get the combined 
contrasts then? As I mentioned, I am a begginer 
so I can try to follow the examples in the manual 
but combining them somehow seems too much for me... Any help is welcome :-)

One more question: I filtered out some bad spots 
and the values are set no NA. Will they be simply 
ignored and only existing values from other array 
replicates will be combined for a given gene? 
Hope so. I know that giving spot weights would be 
better but that's left for the future...


Dr Agnieszka ¯mieñko

Centrum Doskonalosci CENAT
Instytut Chemii Bioorganicznej Polskiej Akademii Nauk
Noskowskiego 12/14
61-704 Poznañ
tel. (61) 8528503 wew. 249
fax: (61) 8520532

Agnieszka Zmienko, Ph.D.

CENAT
Institute of Bioorganic Chemistry
Polish Academy of Sciences
Noskowskiego 12/14
61-704 Poznan, Poland
phone (0048) 61-8528503 ext. 249
fax: (0048) 61-8520532