[BioC] bugs in arrayQuality and marray

[Paquet, Agnes]

Hi Francois,

I have started looking at the problems you reported last week. From a
visual inspection of your image processing output file, I think that the
problem comes from the way Agilent image processing software is managing
the orange packing format. Agilent image analysis software is keeping
the orange packing format in the output file (which is different from
other software package like GenePix) is a way that is not accounted for
in marray. 

The layout of the array in the file looks something like:

Feature Num  Row	Col
      1       1    1
      2       1    3
      3       1    5...

      215     1    429
      216     2    2
      217     2    4
      218     2    6 
       

The real layout of your array should be 215 columns by 206 rows, but
marray is reading 430 columns by 206, creating an object with twice as
many spots, as you reported...

I will forward this problem to Jean Yang, who is the maintainer of
marray, so that she can address this issue.

Thank you very much for reporting this,

Agnes

-----Original Message-----
From: bioconductor-bounces at stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of Francois
Pepin
Sent: Friday, December 02, 2005 4:14 PM
To: BioClist
Subject: [BioC] bugs in arrayQuality and marray

Hi everyone,

(note to moderators: please reject my previous message, I'm resending
with a link to the file instead of an attachment).

I've hit a couple of bugs while trying out arrayQuality (downloaded from
ucsf site, they have an earlier version than bioconductor) for my
Agilent arrays.

read.Agilent, from the marray package, is ignoring the 'path' parameter
when 'fnames' is given:

    if (is.null(fnames)) {
        ...
    }
    else fullnames <- fnames



The readAgilent function, from the arrayQuality package, has problems
with the skip parameter. First of all, it is completely ignored:

   skip <- grep("gMedianSignal", y)[1]

Then lower down, it is incorrectly offset by +1:

    dat <- scan(f, quiet = TRUE, what = h, sep = sep, skip = skip +
        1, quote = quote, ...)

which means that the first gene is skipped (which then cause problems
because dimensions are not of the same).

Then maPrintTip in marray is confused with how Agilent probes are
positioned and seems to think that there are twice as many probes as
there are. Doing a print tip-based normalization doesn't make much sense
in this case (only 1 print tip is present), but it did end up making the
code crash. This can be fixed by specifying norm='l' (if loess
normalization is wanted) in the agQuality.

I think that should be the default in agQuality and change (in
agQuality):
        defs <- list(norm = "p")
to 
        defs <- list(norm = "l")

As a note, the diagnostic spatial images being produced are not correct
either. It produces 4 copies of the image rather than a single one. I've
put at put  the output from a truly horrible array we have where it's
obvious on my website
http://www.mcb.mcgill.ca/~francois/diagPlot.US22502628_251239112296_S01_
A01.png
. I assume this is again due to the awkward way Agilent has to
place its spots.

> sessionInfo()
R version 2.2.0, 2005-10-06, x86_64-redhat-linux-gnu

attached base packages:
[1] "grid"      "tools"     "methods"   "stats"     "graphics"
"grDevices"
[7] "utils"     "datasets"  "base"


other attached packages:
arrayQuality RColorBrewer     gridBase       hexbin   colorspace
convert
     "1.6.2"      "0.2-3"      "0.4-1"      "1.4.0"        "0.9"
"1.4.0"
     Biobase       marray        limma
     "1.8.0"      "1.8.0"      "2.3.7"


Francois

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/bioconductor