Robust Multi-array Average could get negative value or not?
1
0
Entering edit mode
Yang Shi • 0
@ea61ff7a
Last seen 2 days ago
Zheng Zhou

Dear Communities, The series matrix of GSE148160 downloaded directly from GEO database showing negative value. So CEL files were processed by oligo package. The value of affyRaw showed non-negative value (attached figure). But there was negative value after oligo::rma procedure (attached figure). In my opinion, the negative value should not be got after rma only if the data was scale or center to zero further. Could you please explain this to me? Thanks so much!

affyRaw <- oligo::read.celfiles(celFiles)
summary(exprs(affyRaw))
eset <- oligo::rma(object = affyRaw)
summary(exprs(eset))


pd.hg.u219 oligo AffymetrixChip • 177 views
2
Entering edit mode
@james-w-macdonald-5106
Last seen 3 hours ago
United States

After running RMA, your data have been logged, base 2. Any value < 1 will have a negative value. Looking at the data from that GEO submission, it appears the vast majority of the reads have very low expression levels. In fact, it looks completely busted.

> dat <- read.celfiles(list.celfiles(listGzipped = TRUE))
> eset <- rma(dat)
Background correcting
Normalizing
Calculating Expression

> exprs(eset)[rowMeans(exprs(eset)) > 4L,1, drop = FALSE]
GSM4455278_Control1.ga.cel.gz
11715686_s_at                             4.062802
11731735_s_at                             3.961158
11742888_s_at                             5.665980
11753638_s_at                             3.907662
11756405_s_at                             5.621388
11757061_s_at                             5.866299
AFFX-r2-Ec-bioB-3_at                      8.127814
AFFX-r2-Ec-bioB-5_at                      9.264246
AFFX-r2-Ec-bioB-M_at                      9.003103
AFFX-r2-Ec-bioC-3_at                      9.873585
AFFX-r2-Ec-bioC-5_at                      9.984458
AFFX-r2-Ec-bioD-3_at                     11.690204
AFFX-r2-Ec-bioD-5_at                     11.706231
AFFX-r2-P1-cre-3_at                      12.620508
AFFX-r2-P1-cre-5_at                      12.314118
>


Having a total of 6 non-control probesets above 4 (and just barely!) indicates that these arrays are probably complete failures. I am actually a bit surprised that they were submitted, but what do I know? I would not personally use these data for anything.

0
Entering edit mode

Thanks for your quick reply sir! Could you tell me why 4 which you selected here?

0
Entering edit mode

Because it's a small number.

0
Entering edit mode

Got the point sir~ Thanks so much!