[BioC] removing outlier/masked probes and gcrma

Tue Jan 16 08:36:43 CET 2007

> Thanks, Jim, for the thoughts below.  Unfortunately, we are using a
> custom Affy chip design here, so those precomputed ones won't work for
> us.  But we could certainly create a custom CDF file for our chip type
> too...

The package altcdfenvs will let you do so (almost) painlessly.

> .. but it surprises me somewhat that there isn't an alternate solution.
> First, what do people do with an AffyBatch object which was read in
> using the rm.mask option if it can't be used for further analyses?  (Or
> is this a failing in how gcrma specifically deals with NAs?)

Dealing with unknown values (NAs) is quite a general problem, and
although some methods can accommodate their presence, some will not
be able to proceed.
One way is to try to infer what are your missing data: they
are several approaches to pick from (ranging from taking the average to
more elaborated techniques), keeping in mind that
there is probably no magic wand for missing values - otherwise it
would have been built into all methods. Once you guessed what the missing
values could be, you can apply your processing method.

You may also consider something else than gcrma to process your data.

>  And
> second, although custom CDFs would be great for dealing with
> ChipType-specific effects (e.g., SNPs), how do people deal with
> chip-specific effects (e.g., scratches and debris)?

affyPLM is letting you fit chip-effect-at-the-probe-level and the likes,
I think. That can be something to help you.

hoping this helps,

Laurent

>  Just a couple of
> thoughts...  Any additional ideas are welcome, but we'll be pushing
> ahead on custom CDFs in the mean time...
>
> Cheers,
> -andrew
>
>
>
> -----Original Message-----
> From: James W. MacDonald [mailto:jmacdon at med.umich.edu]
> Sent: Saturday, January 13, 2007 6:41 AM
> To: Andrew Su
> Cc: bioconductor at stat.math.ethz.ch
> Subject: Re: [BioC] removing outlier/masked probes and gcrma
>
> Hi Andrew,
>
> Andrew Su wrote:
>> I am attempting to use gcrma on AffyBatch objects which were read in
>> using the "rm.outliers=TRUE" or "rm.mask=TRUE" options (to the
> ReadAffy
>> function).  For example, I put two MOE430 CEL files in the working
>> directory, and here is what I tried:
>>
>>
>>
>>
>>>ab<-ReadAffy(filenames=list.celfiles(),rm.outliers=TRUE)
>>
>>
>>>ai<-compute.affinities(cdfName(ab))
>>
>>
>> .> data<-gcrma(ab,ai)
>>
>> Adjusting for optical effect..Done.
>>
>> Adjusting for non-specific binding.Error in
>> gcrma.bg.transformation.fast(pms, bhat, var.y, k = k) :
>>
>>         NAs are not allowed in subscripted assignments
>
> As you can see, you cannot have any NAs in your data to use gcrma. An
> alternative to this is to use the MBNI cdf/probe packages that have the
> probes with SNPs in the central 15 base pairs removed. Anything in this
> listing with SNP in the name has these probes removed.
>
> http://brainarray.mbni.med.umich.edu/Brainarray/Database/CustomCDF/CDF_d
> ownload_v6.asp
>
> Note that there are some downsides to using these cdfs, mainly that the
> standard errors of your estimates will be highly variable, since the
> probesets for these cdfs are quite variable in size (unlike the stock
> affy chip, where the vast majority have 11 probes).
>
> Best,
>
> Jim
>
>
>>
>>
>>>sessionInfo()
>>
>>
>> Version 2.3.1 (2006-06-01)
>>
>> i386-pc-mingw32
>>
>>
>>
>> attached base packages:
>>
>> [1] "splines"   "tools"     "methods"   "stats"     "graphics"
>> "grDevices"
>>
>> [7] "utils"     "datasets"  "base"
>>
>>
>>
>> other attached packages:
>>
>> mouse4302probe   mouse4302cdf          gcrma    matchprobes
>> affy
>>
>>       "1.10.0"       "1.10.0"        "2.6.0"        "1.4.0"
>> "1.12.2"
>>
>>         affyio        Biobase
>>
>>        "1.0.0"       "1.10.1"
>>
>>
>>
>>
>>
>> I have tried using both R versions 2.3.1 and 2.1.0, and gcrma versions
>> 1.1.4 and 2.6.0, and affy versions 1.12.2 and 1.10.0.  I get a similar
>> error when using the rm.mask=TRUE option.
>>
>>
>>
>> My overall goal is to remove select probes from the analysis (in this
>> case, probes that overlap known polymorphisms).  Any thoughts on how
>> best to do this are most appreciated...
>>
>>
>>
>> Cheers,
>>
>> -andrew
>>
>>
>>
>> --
>>
>> Andrew Su, Ph.D.
>>
>> Genomics Institute of the
>>
>>   Novartis Research Foundation
>>
>> asu at gnf.org
>>
>> Tel: 858-812-1656
>>
>> Fax: 858-812-1630
>>
>> http://web.gnf.org
>>
>>
>>
>>
>> 	[[alternative HTML version deleted]]
>>
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives:
> http://news.gmane.org/gmane.science.biology.informatics.conductor
>
>
> --
> James W. MacDonald
> University of Michigan
> Affymetrix and cDNA Microarray Core
> 1500 E Medical Center Drive
> Ann Arbor MI 48109
> 734-647-5623
>
>
>
> **********************************************************
> Electronic Mail is not secure, may not be read every day, and should not
> be used for urgent or sensitive issues.
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives:
> http://news.gmane.org/gmane.science.biology.informatics.conductor
>
> !DSPAM:45abe36b161401355422449!
>
>
>