hi everyone! i need to perform justRMA and ComBat normalizations at the same time on a dataset that was obtained with affymetrix hg133A and another one from hg133plus. can I do it with these functions even if there will be some more probes on the second one or can't I? if i can't, are there any other functions that allow me to perform normalizations within datasets from partally different chips?

Thanks a lot! Susanna

ps: i'm new here so if the same question was already answered pls feel free to point it out and attach the convo. thx

Technically, yes. You might want to subset the hgu133plus2 array (which includes probesets from both the hgu133a and hgu133b, plus some other probesets). You can easily run justRMA on each data set, and then subset the plus2 array to only include the probesets on the hgu133a array. That's the easy part.

The hard part is then what to do with those data, and how advisable might it be? If one of your groups is on the hgu133a array, and the other is on the hgu133plus2 array, direct comparisons are not a good idea. Any biological differences will be confounded with technical differences, which are likely to be much larger than the biological differences. And ComBat can't tell the difference between technical and biological differences in that situation and will simply remove both.

You might have two relatively identical experiments where each has the two groups you wish to compare, in which case you could use ComBat to remove the batch effects (which will be orthogonal to the biological differences), and then make comparisons. I probably wouldn't do that though. When Affy started out, they didn't provide the kit for the reverse transcriptase step, and instead people bought the kit from a different company. But later Affy did start making the kit, and people mostly seemed to switch to using their kit. This was around the time when Affy started selling the plus2 array, and although Affy told people there were no differences between the kits, that was not actually correct. There were huge differences, and you simply couldn't make any direct comparisons between data processed using the two kits. It is easily possible that the hgu133a data were generated using the original kit, and the plus2 data using the Affy version.

To be safe, I would instead use the GeneMeta package to do a meta-analysis of the summary statistics generated by making the within-array comparisons. But that is only if you have both groups you care about on both arrays. If not, I would advise forgetting about even trying this.