How do microarray manufacturers decide on the number of probe pairs in a probeset?

Question

[Volka]

Hi all, been reading about the technical details of microarrays, and I am just wondering how do companies like Affymetrix choose the number of probe pairs for their probesets? For instance, with their arrays, it is usually 11-20 probe pairs per probeset, how is this decided?

Thanks very much!

[Benn]

This questions seems not to be really about bioinformatics, but about microarray manufacturing. Maybe you can try to get an answer on seqAnswers

Answer 1

[GenoMax]

Prime consideration is to cover as much of the gene as possible and as specifically as possible. One wants to minimize the chance of probes cross-reacting across genes. Affymetrix started using perfect-match/mismatch pair combination for each probe set as an additional measure but as I recall after a certain time many of the array packages were ignoring the mismatch probes since they did not always help in the way they were intended to (it has been a while since I have thought about arrays so my memory is a bit fuzzy).

Answer 2

[Ahill]

I'll add that one consideration driving the number of probe-pairs per probe-set in later Affymetrix designs was the desire to query more transcripts in the same amount of silicon real estate, given the fixed size (in square microns) of each probe feature, which was determined by the resolution of the photolithography process. Very early Affymetrix arrays (or custom arrays) contained up to ~100's of probe-pairs per probe-set, and queried a smaller number of transcripts. Reducing the number of probe pairs per transcript to ~ 22, 20, and later 11 allowed later genome-scale array designs to query more distinct transcripts, with what was considered a modest cost in increased signal variance at the transcript level.