[BioC] Single Channel Approach for Agilent Arrays

[Gaj Stan (BIGCAT)]

Dear Naomi,

Thanks for your initial response.

My intention was to test out the single channel analysis (SCA) approach
on a very simple and easy dataset. 

The reason behind choosing SCA for this dataset was based upon finding a
larger amount of genes that reacted inconsistently (|FC| >1.4 and
opposite direction on both the normal and the dye flip) during my
two-channel analysis compared with the amount of genes that reacted
consistently on both arrays. This could be explained as a possible
dye-effect and I thought that a SCA approach might improve these type of
'weird' results, at least for that specific list of genes. I do
understand that the SCA enables you to do indirect comparisons between
different two-coloured arrays (using contrasts), and that that is not my
main aim here. I'm trying to catch up on the subject, so do correct me
if I'm wrong! (-;

I do have other experimental Agilent data constructed in a more general
loop-design (also dye-flipped) with the same issue at hand. However, if
I use the SCA for that, then I'd still come across my 'pretty basic'
questions mentioned below, i.e. take dye-effect into design-file and
bypassing the 'error' in the intraspotCorrelation function (-:

Best wishes,

   Stan


-----Original Message-----
From: Naomi Altman [mailto:naomi at stat.psu.edu] 
Sent: 03 October 2006 05:13
To: Gaj Stan (BIGCAT); bioconductor at stat.math.ethz.ch
Subject: Re: [BioC] Single Channel Approach for Agilent Arrays

While I always use the single channel approach for loop designs, I 
see no compelling reason to use it for a reference design, which is 
what you have here.
Since you appear to have technical replicates for the dye swap, I 
would use the 2-channel approach, and save the "blocks" for the 
technical replications.

--Naomi

At 09:17 AM 10/2/2006, Gaj Stan (BIGCAT) wrote:
>Dear BioConductor-user,
>
>I've recently tested out Chapter 9 in the Limma user documentation 
>concerning applying single channel approach on Agilent arrays. My 
>experimental setup consists of two different (pooled) food 
>interventions in dye flip using the same control sample (n=4 - For 
>each experiment, 1 array + dye-flip). I assume that my ultimate 
>single-channel design matrix would look like this
>
>         fControl        fFood1          fFood2
>(1)     1       0       0
>(2)     0       1       0
>(3)     0       1       0
>(4)     1       0       0
>(5)     1       0       0
>(6)     0       0       1
>(7)     0       0       1
>(8)     1       0       0
>
>Where:
>Food1_vs_Control.Cy3 (1)
>Food1_vs_Control.Cy5 (2)
>Control_vs_Food1.Cy3 (3)
>Control_vs_Food1.Cy5 (4)
>Food2_vs_Control.Cy3 (5)
>Food2_vs_Control.Cy5 (6)
>Control_vs_Food2.Cy3 (7)
>Control_vs_Food2.Cy5 (8)
>
>Question 1: Am I missing any information here? Because if I do this 
>as suggested in the manual, I get a matrix with some extra (and 
>empty) attributes such as assign (1,1,1), contrasts (NULL) and 
>contrasts$f ("contr.treatment"). Are these attributes necessary for 
>further steps?
>
>Question 2: In my normal two-channel analysis approach I added an 
>extra column in my design file that includes the DyeEffect in the 
>statistical analysis (Ebayes) afterwards. Do I need to include them 
>here as well?
>
>The next part is calculating the intraspotCorrelation, based upon 
>the design file above. When executing, I get the following error: 
>"Missing or Infinite Values found in M or A". This error should 
>indeed occur, since individual spots of bad quality were flagged as 
>NA before normalization, resulting in no M or A value (missing 
>value). Is there a way to omit these spots from the 
>intraspotCorrelation function (By changing them to 0 it might have 
>an effect on the average correlation factor?) or does something else 
>need to be done to solve this?
>
>Any suggestions would be greatly appreciated!
>
>Best wishes,
>
>    Stan
>
>
>---------------------------------------
>Stan Gaj, MSc
>PhD Student
>Dept. of Human Biology / BiGCaT Bioinformatics / Nutrigenomics
Consortium
>PO BOX 616
>UNS 50 - Box 28
>University Maastricht
>6200MD Maastricht
>the Netherlands
>
>Tel:  +31 (0)43 3882913
>Fax: +31 (0)43 3670976
>
>
>_______________________________________________
>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

Naomi S. Altman                                814-865-3791 (voice)
Associate Professor
Dept. of Statistics                              814-863-7114 (fax)
Penn State University                         814-865-1348 (Statistics)
University Park, PA 16802-2111