[BioC] Affy: Present calls in an eset

Crispin Miller CMiller at PICR.man.ac.uk
Wed Oct 8 16:54:10 MEST 2003


MAS5.0 gives you two numbers, the expression level for each gene, and a
detection p-value that tells you how confident MAS5.0 is that it got the
expression level right. This p-value is then split into three categories
(simply by thresholding) to give you Present, Marginal or Absent calls.
Generally, thresholding by intensity does something similar to using the
detection p-value, but not always, for every gene, in every situation.
Keeping these two concepts distinct (accuracy and confidence), is IMHO
not a bad idea (we can of course argue about how best to calculate
metrics for these :-)). After all, this seems to be what we are doing
when we generate not only a fold-change but also a p-score across
replicates for each gene in a study.


> -----Original Message-----
> From: Phguardiol at aol.com [mailto:Phguardiol at aol.com] 
> Sent: 08 October 2003 15:29
> To: Arne.Muller at aventis.com; bioconductor at stat.math.ethz.ch
> Subject: Re: [BioC] Affy: Present calls in an eset
> Hi,
> my understading ot the A/P calls from Affy MAS5 is that they 
> represent a way 
> to "evaluate" the background signal. 
> In that way you could have a high background, ie, a high MisM 
> signal, and 
> despite the PM signal is identical, for instance 500, the 
> this gene is going to 
> be considered as Absent, which could be that your gene is not 
> expressed, but 
> could also be due to the design of the probe set rather than 
> to the fact that 
> the gene is not expressed....
> Detecting if a gene is expressed or not with Affy chips is 
> for me almost 
> another question, a difficult one ! my threshold is 70 (2^RMA 
> value) but it is 
> likely that at this value some genes are expressed it is hust 
> that below tis 
> value I consider, arbitrarly, that it is difficult to 
> validate something because 
> of the noise.
> Anybody can correct me if I m wrong with all this since I  
> still new in ths 
> field as well !
> Hope it helps
> Philippe  
> 	[[alternative HTML version deleted]]
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch 
> https://www.stat.math.ethz.ch/mailman/listinfo> /bioconductor

This email is confidential and intended solely for the use o...{{dropped}}

More information about the Bioconductor mailing list