Hi Chris,

You will spend plenty of time combining directly Gene ST 1.0 and 430 2.0 data and this will go nowhere.

Even a comparison of the same type of array processed in different labs is very problematic due to technical variation. Irizarry reduces expression measurements of different labs (on identical arrays) to a yes/no question to work around the technical differences (http://nar.oxfordjournals.org/content/39/suppl_1/D1011.full).

430 2.0 arrays have all probes in the 3' end of transcripts, Gene ST 1.0 arrays have the probes spread across the transcript. Processing of both array types also uses very different chemistry.

This is what could be done (in my opinion): If there is the same kind of treatment in the CNS samples and in the PNS samples, you can quantify differential expression within each group. Afterwards, you could compare differential expression between groups. This eliminates the probe inherent effects.

To obtain high quality data, you have to isolate RNA from both cell types and process them together on the same array.

Sorry for the bad news,

Herbert

If the probe(set) sequences of a rather large subset of the genes is similar (same) on both arrays then you can focus on this subset and do a qunatile-normalization between the arrays. You can then use this normalized data to check for differences in gene expression. If your gene(s) of interest is/are not part of this common subset then you have a problem.

If the probe(set) sequences are mostly all different, you will get different signals for the genes. One can still normalize but this will not alter the inherent "distortion" of the signal distribution between the arrays (in other words: if you would hybridize the same sample on both arrays you will gind a huge proportion beeing appaerantly quite strongly differentially expressed - whether or not you normalize, since you are comparing "apples and peaches"). The genes being very very drastically differentially expressed should still be identifiable, though, but I have no idea what a reasonable minimum difference this would be and if it is what you are looking for.

Dear Chris,

Quantile normalization is a great approach for making two distributions similar, just as Jochen has pointed out.

You may also look into the z-score normalization of microarray expression, which also allows the comparison of microarray data independent of the original hybridization intensities.

The Z score transformation approach fcorrects data internally within a single hybridization and hybridization values for individual genes are expressed as a unit of standard deviation from the normalized mean of zero. Correction is done before sample-to-sample comparison, and is therefore comparison-independent

Have a look at this paper for more info.

Article Analysis of Microarray Data Using Z Score Transformation

This is great! Thanks very much for the rapid responses. I clearly have a lot of reading to do, and some programming to learn, but now I know it will actually be worthwhile!

Thanks!

Chris, I've not used this at all yet, but there is one pipeline designed just for meta analysis of different Affy arrays - MAAMD (http://www.biokepler.org/use_cases/maamd-workflow-standardize-meta-analyses-affymetrix-microarray-data).  It might be interesting to check into it, or at least some of the tools used in the pipeline (all open source I believe).

Hi Chris,

You will spend plenty of time combining directly Gene ST 1.0 and 430 2.0 data and this will go nowhere.

Even a comparison of the same type of array processed in different labs is very problematic due to technical variation. Irizarry reduces expression measurements of different labs (on identical arrays) to a yes/no question to work around the technical differences (http://nar.oxfordjournals.org/content/39/suppl_1/D1011.full).

430 2.0 arrays have all probes in the 3' end of transcripts, Gene ST 1.0 arrays have the probes spread across the transcript. Processing of both array types also uses very different chemistry.

This is what could be done (in my opinion): If there is the same kind of treatment in the CNS samples and in the PNS samples, you can quantify differential expression within each group. Afterwards, you could compare differential expression between groups. This eliminates the probe inherent effects.

To obtain high quality data, you have to isolate RNA from both cell types and process them together on the same array.

Sorry for the bad news,

Herbert