I have data from two different Affymetrix platforms (Affymetrix hugene 2.1st and Affymetrix U133a). I’d like to compare the mean expression of the two samples (sample 1: Affymetrix hugene 2.1st, sample 2: Affymetrix U133a).
Below you can find boxplots of the raw data (four boxplots of each platform shown).
Are there any guidelines for merging and preprocessing the data? Can you recommend a bioconductor package for combining arrays from different platforms (bioconductor version 3.7)?
You shouldn't try to merge data between Affy arrays. In the first place, what is being measured is different between a 3'-biased and a random primer array. In addition, both of those arrays have multiple probesets for many of the genes being measured, so it's not so simple to say which probesets you should consider 'the same' in the two arrays.
Back in the day Affy used to recommend an IVT kit made by Enzo for their 3'-biased arrays. When they switched to their own IVT kit, they claimed that the results were comparable, which they were, for some definition of comparable. But what they didn't tell people was that you couldn't actually mix data that were processed using a combination of the Enzo and Affy IVT kits! Just changing one step in the process made the data sufficiently different that you could no longer just combine things unless you had orthogonal batches.
Anyway, when confronted with this type of scenario I usually go with the GeneMeta package, although there are probably others you could try as well.
