R Gcrma For Chipset Without Mm Probe Intensity

Question

[AngryBird]

Hi, I am trying to perform GCRMA for CEL files from mogene10stv1 microarray. This microarray is only designed for PM intensities. How should I proceed?

Thanks.

Answer 1

[Neilfws]

This question comes up quite frequently on R and Bioconductor mailing lists. See for example:

• Gcrma And Chip Without Mm Values
• Tiling array application
• gcrma problem while processing HuGene-1_0-st-v1 genechip from Affymetrix

I think the best answer is in the last of those 3 links and I'll quote it here:

Note however that gcrma() has an 'NCprobe' argument that you can use to specify an index of negative control probes. This is a non-trivial thing to do, and may be beyond your abilities if you are very new to R and BioC.

To get the index of these probes, you will need to decide which probes are negative control probes, and then you can use the indexProbes() function, passing a character vector of the negative control probes to the genenames argument. This will return a list of indices for each probeset that you can unlist() prior to feeding in to gcrma().

Or you could just stick with rma(), like the vast majority of people do.

In summary: GCRMA was designed for use with MM probes. If you don't have them, better to use RMA.

[AngryBird]

Yes, I agree this question comes up frequently but isn't answered to a level of detail, which can solve the problem.

How do you determine which probes are negative controls referred to you by "passing a character vector of the negative control probes", what is the criteria? Is it experiment dependent or a standard.

Thanks

[Neilfws]

Well, you need to know which probesets the manufacturer of that chip considers to be negative controls, which is probably somewhere in the product literature. However, in this case I'd forget about gcrma and just use rma, as suggested. Frankly, the difference in normalization methods will be slight and insignificant to downstream analyses.