[BioC] RNA degradation plot with oligo package GeneFeatureSet objects

heyi xiao xiaoheyiyh at yahoo.com
Sat Aug 17 00:26:41 CEST 2013


Hi Jim,
You mean, I can’t use makecdfenv package to build CDF package from PGF and CLF files?
I will definitely check these other plots before decide on quality issue.
Thanks!
Heyi

--------------------------------------------
On Fri, 8/16/13, James W. MacDonald <jmacdon at uw.edu> wrote:

 Subject: Re: [BioC] RNA degradation plot with oligo package GeneFeatureSet objects

 Cc: bioconductor at r-project.org
 Date: Friday, August 16, 2013, 3:59 PM

 Hi Heyi,

 No you can't use the affy package for that chip type. As an
 aside, I 
 have never found an array that I wanted to exclude based on
 the RNA 
 degradation plot that I hadn't already decided to exclude
 based on 
 either a PCA plot, a density plot of the raw data, or a NUSE
 or RLE 
 plot. In other words, I think there are much better ways to
 find outlier 
 plots than the degradation plot.


 Best,

 Jim



 On 8/16/2013 3:33 PM, heyi xiao wrote:
 > Hi Jim,
 > Thanks for the informative notes. I really learned
[[elided Yahoo spam]]
 > I see you what mean. But I only use RNA degradation as
 a quality assessment tool. I am less interested in estimate
 exactly how much RNA degradation happened in the RNA
 molecules in one sample/array, I am more interested in the
 different degradation patterns seen across different
 samples. Normally degradation curves for different samples
 stack together consistently and nicely. Even with the newer
 generation Affy arrays, an outlier degradation curve always
 suggest some quality issue, mostly likely RNA degradation.
 Such RNA degradation curves together with other quality
 check help me either kick out the problematic samples or
 have them redone.
 > BTW, currently only oligo package seems to work with
 the new Ovine Gene 1.1 ST array, for which I don’t see an
 CDF package in bioconductor as other affy chip types.
 Therefore, I can’t go with affy and other packages which
 provide RNA degradation plots. Can I use makecdfenv package
 to build CDF package from PGF and CLF files?
 > Heyi
 >
 > --------------------------------------------
 > On Fri, 8/16/13, James W. MacDonald<jmacdon at uw.edu> 
 wrote:
 >
 >   Subject: Re: [BioC] RNA degradation
 plot with oligo package GeneFeatureSet objects

 >   Cc: bioconductor at r-project.org
 >   Date: Friday, August 16, 2013, 12:52
 PM
 >
 >   Hi Heyi,
 >
 >   On 8/14/2013 4:47 PM, heyi xiao
 wrote:
 >   >  Hi all,
 >   >  In affy package, I can use
 AffyRNAdeg and
 >   plotAffyRNAdeg to plot and check RNA
 degradation. Is there
 >   any way to do so in oligo package for
 GeneFeatureSet,which
 >   is equivalent to AffyBatch in affy
 package. I look at the
 >   GeneFeatureSet and AffyBatch, they
 quite similar. But not
 >   sure what can be done here. I can
 either modify AffyRNAdeg
 >   and plotAffyRNAdeg functions to fit
 them for GeneFeatureSet,
 >   or I can convert GeneFeatureSet to
 AffyBatch and use the
 >   affy package degradation functions.
 Any suggestions would be
 >   highly appreciated.
 >
 >   While I suppose you could
 hypothetically do the conversion,
 >   I wonder if it makes conceptual
 sense.
 >
 >   The 3'-biased Affy arrays were all
 based off an oligo-dT
 >   primer that was used to convert mRNA
 to cDNA, so the reverse
 >   transcription proceeded from the 3'
 end of the mRNA, always.
 >   In this case you can wonder about two
 things. First, how far
 >   did the RT step proceed? Did you in
 general get good RT all
 >   the way to the most 5' of the probes
 in the probesets?
 >
 >   Second, since we were using the polyA
 tail at the 3' end, by
 >   definition the mRNA wasn't degraded
 from the 3' end.
 >   However, it might have had more or
 less extensive
 >   degradation from the 5' end, so the RT
 may have gone to
 >   completion, but the degradation had
 proceeded past the most
 >   5' probes.
 >
 >   So both things are confounded, as we
 cannot distinguish RT
 >   that didn't proceed too far from
 highly degraded mRNA, but
 >   no matter. What we could do is say how
 much signal we were
 >   getting from the more 5' probes, and
 decide if we wanted to
 >   do something about that (like only use
 the first 8 probes or
 >   whatever).
 >
 >   For the newer generation of Affy
 arrays, we use a random
 >   primer, so the RT step proceeds from a
 random point in the
 >   transcript and proceeds towards the 5'
 end (at least I think
 >   it is still directional). Since the RT
 no longer starts from
 >   one end of the transcript, it is no
 longer clear what
 >   differential amounts of probe signal
 would actually
 >   signify.
 >
 >   In addition, with the newer generation
 of Affy arrays, we
 >   can collapse the probes into different
 probesets, depending
 >   on what we are trying to measure
 (e.g., you can try to
 >   measure expression at the exon level
 or the transcript
 >   level).
 >
 >   I think trying to do this would be
 more difficult than it
 >   would be worth, especially given that
 I don't know what you
 >   would do if you were to decide there
 had been degradation.
 >
 >   Best,
 >
 >   Jim
 >
 >
 >   >  Heyi
 >   >
 >   > 
 _______________________________________________
 >   >  Bioconductor mailing list
 >   >  Bioconductor at r-project.org
 >   >  https://stat.ethz.ch/mailman/listinfo/bioconductor
 >   >  Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
 >
 >   -- James W. MacDonald, M.S.
 >   Biostatistician
 >   University of Washington
 >   Environmental and Occupational Health
 Sciences
 >   4225 Roosevelt Way NE, # 100
 >   Seattle WA 98105-6099
 >
 >

 -- 
 James W. MacDonald, M.S.
 Biostatistician
 University of Washington
 Environmental and Occupational Health Sciences
 4225 Roosevelt Way NE, # 100
 Seattle WA 98105-6099



More information about the Bioconductor mailing list