[BioC] Are individual channels' intensities comparable across cDNA arrays?

[Sean Davis]

On Wed, Oct 15, 2008 at 7:14 PM, Seungwoo Hwang <swhwang10 at yahoo.com> wrote:
> Dear all,
>
> I received the following request. I'm not so sure about it and I would like to receive your expert opinion on this.
>
> The platform is cDNA array with two experimental series.
> - On series 1 chips, tumor sample from a patient (Cy5) and normal sample from pooled healthy individuals (Cy3) were co-hybridized.
> - On series 2 chips, cirrhosis sample from that patient (Cy5) and and normal sample from pooled healthy individuals (Cy3) were co-hybridized.
> - This hybridization scheme was utilized because they wanted to make the following three comparisons: tumor vs normal, cirrhosis vs normal, and tumor vs cirrhosis.
>
> I normalize by print-tip loess for within array and scale method for between arrays using limma. After that, I usually work on normalized M values for differential expression analysis.
>
> The request that I received, however, was to obtain tumor sample intensities (Cy5 on series 1 chips) and cirrhosis sample intensities (Cy5 on series 2 chips) so that they can do some comparative analyses with their protein data.
>
> I know that I am able to obtain those values using RG.MA(). What I am not sure is, will those two individual channels' intensity values become comparable with each other? By comparable, I mean, when a gene's M is 1 in series 1 chips and 2 in series 2 chips, I am not sure whether that 2-fold change (M of series 2/ M of series 1) will be shown as 2-fold change of individual Cy5 intensities (R of series 2/ R of series 1) also.
>
> I do not think that the ratio of individual Cy5 intensities do not become 2-fold in this example, because normalization works on M values, which would make M values comparable across arrays but not individual Cy5 intensities. Is this reasoning correct?
>

If you want to compare intensities, you will probably need to do some
between-array normalization.  The limma user guide has some examples
of ways to do this keeping in mind that you do have two channels of
data.

Sean