Trim down to 22011 pairs for Human Exon

[smithjessk]

Look at the basicRMA function in oligo

[smithjessk]

https://github.com/Bioconductor-mirror/oligo/blob/321950c77e9c6dc9f9ae86b296cf9215e2cb0f28/R/methods-summarization.R#L117

Number of unique fsetids in the perfect match probes = 22011.
0 through Count(fsetid) - 1 are split into nPn groups, each labeled with the corresponding entry in unique(pnVec). Note that, before this method, pnVec is sorted by fsetid.

Example:

$`2365391`
 [1] 30957 30958 30959 30960 30961 30962 30963 30964 30965 30966 30967 30968
[13] 30969 30970 30971 30972 30973 30974 30975 30976 30977 30978 30979 30980
[25] 30981 30982 30983 30984

$`2365496`
 [1] 30985 30986 30987 30988 30989 30990 30991 30992 30993 30994 30995 30996
[13] 30997 30998 30999 31000 31001 31002 31003 31004 31005 31006 31007 31008
[25] 31009 31010 31011 31012 31013 31014 31015 31016 31017 31018 31019 31020
[37] 31021 31022 31023 31024 31025 31026 31027 31028 31029 31030 31031 31032
[49] 31033 31034 31035 31036 31037 31038

[smithjessk]

The reduction to 22011 is done for each chip. One entry per chip is produced for each unique fsetid. The value of this entry is (I believe) the arithmetic mean, recalculated for each chip, of all the fids that correspond to that fsetid.

[smithjessk]

An easy way to do this is to do the mean inside the MedianPolish GIST. It needs a state that holds the correspondence between fids and fsetids.
a) If we can have a map this would be easy. Could also recreate the map with a matrix and an array
b) could also have a array where the value of entry i is indicates that rows with indices from entry i - 1 to entry i correspond to the fsetid array entry i