[BioC] how to annotate Illumina HumanHT-12 v3 chips?

[Kevin R. Coombes]

My opinion: best = largest IQR (as a first approximation). Some probes 
are broken, and never give meaningful values; they tend to have small IQR.

Keep in mind that this still does not deal with the fact that 
(regardless of what we or the companies designing the probes) might like 
to think, some probes simply do not measure what they are supposed to 
measure.

Here is an interesting (in the ambiguous sense of "may you live in 
interesting times") exercise.  Pick any data set using Affymetrix U133 
arrays with about 100 samples.  Extract the data for all probe-sets that 
target fibronectin.  Construct a pairs plot of the extracted data.  The 
ask yourself (1) if you really believe all probes are measuring the same 
thing and (2) which probes you want want to use as your best guess at a 
reasonable measure of fibronectin expression.

     Kevin

On 3/6/2013 10:20 AM, Steve Lianoglou wrote:
> Hi,
>
> For sake of discussion/pedagogy, when you say:
>
>> Sometimes you have to collapse the data to one probe per-gene, and for
>> this I tend to pick the "best" probe for each gene by using a  measure
>> such as IQR across the dataset.
> In this scenario, which probe is "best" -- is it the one with the
> largest IQR, or the smallest?
>
> Assuming both probes map uniquely to the genome, both of which are (as
> far as we know) "probing" the same gene of interest, what are some
> "sound" reasons to pick one over the other?
>
> -steve
>