[BioC] intraspotcorrelation() in individual channel analysis

Gordon K Smyth smyth at wehi.EDU.AU
Mon Jun 8 03:20:21 CEST 2009


Dear Thierry,

The person who said that the intraspot correlation should be largish was 
almost certainly me, although I would probably have suggested in the range 
0.5-0.9 rather than close to one.  I told you two weeks ago that this 
advice didn't apply to your current platform, and advised you to go ahead 
with the analysis despite your smallish correlation.  So what are you 
"struggling" with?

A low correlation could conceivably be caused by one or more arrays in 
which one channel is poor quality while the other is fine.  The usual 
plots (MA-plots etc) will help you check for this.  Or it might conversely 
be because your arrays are such high quality that technical variability is 
very small compared to biological.  Or it might be because biological 
variability is very high in your data.  Only the first possibility is a 
problem.  If the QC plots look fine, why not go ahead?

If you want to discuss further, please keep the discussion on this list.

Best wishes
Gordon

> Date: Fri, 05 Jun 2009 12:03:28 -0400
> From: Naomi Altman <naomi at stat.psu.edu>
> Subject: Re: [BioC] intraspotcorrelation() in individual channel
> 	analysis
> To: thierry.janssens at ecology.falw.vu.nl
> Cc: bioconductor at stat.math.ethz.ch
>
> Dear Dr. Janssens,
>
> In the studies we did using custom Agilent arrays, the intraspot
> correlation is about 0.3.  The intraspot correlation is residual
> correlation, so it is adjusted for the mean.  This makes it much
> lower then the correlation between the channels that you might compute naively.
>
> Naomi
>
>
> At 07:04 AM 6/5/2009, you wrote:
>> Dear Dr. Altman,
>>
>> on the BioC forum I have noticed you are an expert and frequent user
>> of the anaysis of indivdual channel data of minro-arrays.
>> I am struggling already for a couple of weeks with a very low
>> (alomost zero) consenus intraspotcorrelation in limma. I am afraid
>> that this affects the testing procedure. I read on the forum that it
>> should approach one.
>> I am using custom made 8x15k agilent arrays in an interwoven loop
>> design without reference.
>>
>> Do you know what can be the reason for my low intraspotcorrelations?
>>
>> thank you and kind regards,
>>
>> Thierry Janssens
>>
>> --
>> Thierry K.S. Janssens
>> Vrije Universiteit Amsterdam
>> Faculty of Earth and Life Sciences
>> Institute of Ecological Science
>> Department of Animal Ecology,
>> De Boelelaan 1085
>> 1081 HV AMSTERDAM, The Netherlands
>> Phone: +31 (0)20-5989147
>> Fax: +31 (0)20-5987123
>> thierry.janssens at ecology.falw.vu.nl
>>
>>
>
> Naomi S. Altman                                814-865-3791 (voice)
> Associate Professor
> Dept. of Statistics                              814-863-7114 (fax)
> Penn State University                         814-865-1348 (Statistics)
> University Park, PA 16802-2111



More information about the Bioconductor mailing list