expression matrix from two dyes array

Question

[acamargo]

Hello,

Can anybody tell me how to extract a gene's expression profile from a two dye array data set? I have 10 files from the experiment, each file correspond to an array and each array has two dyes hybridized (cy3 and cy4) and each dye correspond to a sample.

Thanks,
AVC

Answer 1

[Sean Davis]

You mention R. Take a look at the limma user guide. It covers both direct and common reference designs.

[acamargo]

No it doesn't. I've used limma many time before with single channel arrays, it's quite simple to do what I want as one array corresponds to one sample only (at less we're not pooled samples) unlike two dyes arrays where an array (file) corresponds to two samples. Could you please be more specific with your answer?

Thanks a lot

[Sean Davis]

See chapters entitled:

• Two-Color Experiments with a Common Reference
• Direct Two-Color Experimental Designs
• Separate Channel Analysis of Two-Color Data

If those don't cover your experimental setup, perhaps you can give more specifics about your experimental design.

[acamargo]

Thanks for answering so quickly.

I know how to fit the model with two dyes arrays. What I really don't know is how to do, is to extract an expression profile matrix for a given gene. For example:

filename     cy3      cy5
file1.txt    time1    time2
file2.txt    time3    time4

The expression profile will be then:

Genename    time1    time2    time3    time4
myc2        x        x        x        x

This is what I don't know how to do it.

Thanks a lot again

[Sean Davis]

I do not have a good general answer for how to munge the data. If you know how to read in the data (again, limma has lots of approaches to do so), perhaps you can post the code and we can start from there.

[acamargo]

ok, thanks. Seidel has pointed a way to start.

Answer 2

[seidel]

You'll have to build your own matrix, extracting the value for that gene from each channel of the array. The problem with this, is one of normalization. You'd have to explore the difference between normalization the data as a two-channel array, then extracting out the normalized values, versus treating each channel as a single channel array and normalizing that way (which is likely problematic, since the effect of the second dye would not be known). It sounds messy.