[BioC] Most stable gene pairs in array experiment

Michael Dondrup Michael.Dondrup at bccs.uib.no
Tue Oct 20 15:31:32 CEST 2009


Hi,
one can of course compute a lot of things the question is if that  
makes sense. The term 'stability' is as far as i know not well defined
in the statistical sense, however there is a definition of numerical  
stability for algorithms.
http://en.wikipedia.org/wiki/Numerical_stability

Therefor, use of 'stability' in this context is not too helpful. If  
you have multivariate random variables  (genesA,...) you can of course  
define, that
you search for the minimum of of the variance estimate of the  
difference (e.g. var(geneA - geneB)). Thereby creating a new random  
variable from the difference of two
variables Z= X-Y. If application of the variance estimate makes sense  
here, I cannot know.

If you need gene pairs that somewhat 'always react in the way', e.g.  
for qPCR correlation might be a more appropriate concept, as mentioned  
by others, and it's well defined.

If you look only at the difference between gene pairs, the result can  
be influenced by the individual variance of both genes, even if they  
otherwise have the same behaviour under each condition.
Furthermore, one could possibly  apply a correlation test (corr.test)  
and select those genes which are most significant. But this will  
always depend on the desired application which is not clear to me.
One could also choose 'housekeeping' genes by prior knowledge. Maybe  
there are even better methods for selecting controls for qPCR  
experiments from other data in the literature?

Hope that helps even though it's maybe a bit off-topic.

Michael

Am 20.10.2009 um 14:30 schrieb Marcelo Laia:

> 2009/10/20 Michael Dondrup <Michael.Dondrup at bccs.uib.no>:
>> Hi,
>>
>> you can  try something like this or use two for loops:
>>
>>
>>> apply (mygenes, 1 , function(row) { apply (mygenes, 1, function(x) {
>>> var(row-x)  } ) } )
>>
>>        geneA   geneB   geneC   geneD   geneE
>> geneA 0.00000 0.04108 0.06397 0.12217 0.08233
>> geneB 0.04108 0.00000 0.15543 0.12807 0.09365
>> geneC 0.06397 0.15543 0.00000 0.08628 0.08903
>> geneD 0.12217 0.12807 0.08628 0.00000 0.01517
>> geneE 0.08233 0.09365 0.08903 0.01517 0.00000
>>
>>
>> Cheers
>> Michael
>
> Hi,
>
> I am following the discussion and I'm finding very interesting.  
> Congratulations!
>
> After this, I could compare the two genes, two-by-two, and I could
> conclude that the pair with minor variance are the two most stable
> genes of all?
>
> Is this genes appropriated for qPCR internal control? Or am I totally
> wrong here?
>
> Thank you very much!
>
> -- 
> Marcelo Luiz de Laia
> Universidade do Estado de Santa Catarina
> UDESC - www.cav.udesc.br
> Lages - SC - Brazil
> Linux user number 487797
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

Michael Dondrup, Ph.D.
Bergen Center for Computational Science
Computational Biology Unit
Unifob AS - Thormøhlensgate 55, N-5008 Bergen, Norway



More information about the Bioconductor mailing list