Microarray data analysis for different chip.

Question

[raya.girish]

Hi. I have test and control samples obtained from different chip file. Test is obtain from u133a2 file and control is from HuGene 2-0.1v2 There are 76 treated samples (u133a2)and 2 control samples(HuGene chip) I want to perform all 76 with 2 control Gene expression analysis using package like affy and limma . How can I do this type of analysis. I agree that the two samples have different probeset. But I don't know how to deal with such data.

Please help me

[Kevin Blighe]

Hi, it would be wrong of me to comment on how you could technically perform this experiment without first commenting on the experimental setup: you only have 76 test samples versus 2 controls? Is there any reason for this huge imbalance? You will struggle to publish results from this study. Is it a pilot study for a grant? - even in this case, the results would not be credible.

Perhaps you can reassure us by providing some background and context.

[raya.girish]

It just a pilot study

[Kevin Blighe]

Okay, and this is a re-analysis of some previously published data, right?

The experimental setup is just too fraught with bias for the results to be meaningful. Apart from the huge imbalance in comparing 76 cases versus 2 controls, you also have the following issues:

• different microarray chip versions (U133 series versus HuGene)
• many probes [most likely] differ (as you implied)
• all of your cases are one chip version, whilst the controls the other; thus, even if you treated chip version as a covariate in your design model, you'd then be eliminating the differences between cases and controls

My honest advice is to just ignore the controls and to just do some simple analysis with the 76 U133 cases. You could do clustering to see how different cases group together, and also identify very high and very low expressed genes

Edit: You could also process the 2 HuGene samples on their own and do the same simple analyses as per the 76 U133 samples, and then manually compare results. I think that it would reflect negatively if you tried to do an actual differential expression analysis here.