Dear communities,

I want to compute P/A calls at gene-level probesets so that it can be utilized for removing unexpressed probe after rma (target = core). As far as I'm concerned, oligo::paCalls can not handle this. There seems a package "xps" can fix that, which unfortunately could not be installed at Windows 64-bit. Is there some other methods could filter the unexpressed probes? Any advice would be appreciated!

eset <- oligo::rma(object = affyRaw, target = 'core')
xpa <- oligo::paCalls(object = affyRaw, method = 'PSDABG')
xpa <- oligo::paCalls(object = affyRaw, method = 'DABG')

There is no method in oligo to do P/A calls at the transcript level. There are negative control probesets on the array that you could use to estimate the expression of an unexpressed probeset, but unfortunately the binding of an Affymetrix probe is very highly dependent on the GC content of the probe. You can fish out the antigenomic probes and plot the binding as a function of the GC content, and it's apparent from that plot that as the GC content goes up, the binding goes up rapidly as well, and once you hit maybe 50-60% GC content, you will get high binding even though there is nothing complementary in your sample.

Back when I analyzed these data, I almost never filtered. In my opinion it's too hard to do well, and even if you were able to do so, the gains are far outweighed by the work it took to do the filtering.