Hi.
On Wed, Sep 7, 2011 at 10:53 AM, Richard Beyer <rpbe...@gmail.com> wrote:
> Hi,
>
> I'm hoping someone can give me a little guidance for an unusual
> normalization need. I have 9 affy human ST arrays that have varying amounts
> of RNA, so I can't use the standard quantile normalizaton approach. The wet
> lab process that involved the polya spike-ins was done in such a way as to
> ensure the same amount of the polya spike-ins on each array. These probesets
> are:
>
> Probe.Set.ID mRna...Description total_probes
> category
> 7893306 AFFX-Bs-thr_st, polya_spike 1160
> control->affx
> 7894584 AFFX-Bs-lys_st, polya_spike 567
> control->affx
> 7894611 AFFX-Bs-dap_st, polya_spike 1189
> control->affx
> 7895139 AFFX-Bs-phe_st, polya_spike 844
> control->affx
>
> I have done this successfully before using the 3' IVT yeast arrays, but on
> those there were many more probesets, so I just ran expresso without the
> quantile normalization step, and then did a weighted cyclic loess using the
> polya-spike-ins.
>
> I see from the total_probes counts for these 4 probesets on the human ST
> array that maybe I could use a weighted quantile normalization where I set
> the weights = 0 for all the probes except the polya_spike probes. So my
> thought would be to do a RmaBackgroundCorrection() on all probes, then a
> weighted QuantileNormalization() step, then followed with a RmaPlm() call,
> as is done in Mark Robinson's doEverything() functions (which I use all the
> time).
>
> If such an approach is advisable, would someone point out to me how to
> identify the polya_spike probes, so I could then do some weighted
> QuantileNormalization() or, if that's not possible, maybe a cyclic loess
> step?
The problem is that it is not possible (neither in theory I think) to
do a weighted normalization based on the *rank-based* quantile
normalization (this is the one everyone thinks of when saying QN).
There are other ways than a rank-based approach, e.g. what I call
quantile-spline-based QN, cf. help("normalizeQuantileSpline.numeric",
package="aroma.light") versus help("normalizeQuantileRank.numeric",
package="aroma.light").
QuantileNormalization in aroma.affymetrix utilizes the rank-based one
and there is no easy way to change it to use the spline-based one. It
would need to be rewritten. I'll add it to the todo list, but I'm
quite sure it won't happen before 2012. If you're willing to look
into it yourself, have a look at the DChipQuantileNormalization class.
It utilizes normalizeQuantileSpline() and uses zero-one weights, but
the weights are not exposed in the API but only used internally.
There is no implementation of cyclic loess in aroma.affymetrix.
/Henrik
>
> Thanks very much for any help or pointers,
> Dick
> *******************************************************************************
> Richard P. Beyer, Ph.D. University of Washington
> Tel.:(206) 616 7378 Env. & Occ. Health Sci. , Box 354695
> Fax: (206) 685 4696 4225 Roosevelt Way NE, # 100
> Seattle, WA 98105-6099
> http://depts.washington.edu/ceeh/members_fc_IEHSFC.html
> http://staff.washington.edu/dbeyer
> *******************************************************************************
>
> --
> When reporting problems on aroma.affymetrix, make sure 1) to run the latest
> version of the package, 2) to report the output of sessionInfo() and
> traceback(), and 3) to post a complete code example.
>
>
> You received this message because you are subscribed to the Google Groups
> "aroma.affymetrix" group with website http://www.aroma-project.org/.
> To post to this group, send email to aroma-affymetrix@googlegroups.com
> To unsubscribe and other options, go to http://www.aroma-project.org/forum/
>
--
When reporting problems on aroma.affymetrix, make sure 1) to run the latest
version of the package, 2) to report the output of sessionInfo() and
traceback(), and 3) to post a complete code example.
You received this message because you are subscribed to the Google Groups
"aroma.affymetrix" group with website http://www.aroma-project.org/.
To post to this group, send email to aroma-affymetrix@googlegroups.com
To unsubscribe and other options, go to http://www.aroma-project.org/forum/
