[BioC] SCAN.UPC vs NUSE

[Steve Piccolo]

Hi Leif,

Thanks for your email. We have not yet done a formal evaluation to address
how well SCAN.UPC can address artifacts and degraded hybridization, etc.
So for the time being we would recommend that you use NUSE and/or other
quality-control tests before applying SCAN.UPC. However, when summarizing
at the gene/probeset level, we use a 10% trimmed mean, so this may be
adequate in many cases for excluding outlier probes due to streaks,
blotches, etc.

We have considered to use the signal-to-noise ratio or some other metric
that can be derived from the SCAN.UPC calculations as a way to measure
sample quality. But so far we haven't implemented that.

If by chance you decide to do a formal evaluation along these lines, we'd
be happy to discuss it with you.

Thanks,
-Steve



On 9/25/13 4:00 AM, "bioconductor-request at r-project.org"
<bioconductor-request at r-project.org> wrote:

>Message: 20
>Date: Tue, 24 Sep 2013 07:55:34 -0500
>From: "Leif Peterson" <leifepeterson at sbcglobal.net>
>To: <bioconductor at r-project.org>
>Subject: [BioC] SCAN.UPC vs NUSE
>Message-ID: <001801ceb925$5cdd0440$16970cc0$@sbcglobal.net>
>Content-Type: text/plain
>
>Prior to using SCAN.UPC and ComBat, we originally ran NUSE (in AffyPLM)
>and
>removed Affy Hu-133A arrays for which the NUSE criterion was > 1.05.   We
>are wondering whether use of SCAN.UPC without pre-filtering by NUSE would
>be
>sufficient?  Also, regarding artifacts and degraded hybridization regions
>on
>chips (visible in AffyPLM Residual plots in the form of streaks, blotches,
>hot spots, etc.),  we are wondering what SCAN.UPC would do with such
>perturbations?  
>
>
>Leif Peterson
>
>HMRI, Houston