[BioC] question on limma package

J.delasHeras at ed.ac.uk J.delasHeras at ed.ac.uk
Wed Jul 18 17:06:27 CEST 2007


I am not sure B>0 means much, since the B statistic depends on a prior  
knowledge of the % of expected DE genes. At least I see very widely  
different B values for my top genes in different types of experiments.  
I prefer to cut by p value (adjusted p value, by Bonferroni-Hochberg  
-the default- gives you the FDR). I think that's a better starting  
point.

If you have a lot of DE genes that way, do you have a reason to  
believe that they are too many?
If you have too many genes, and you use an FDR of 0.05... you have an  
instant *estimate* of expected false positives in that list (5%). So  
when an experiment gives a lot of genes as DE, maybe you can be more  
strict and reduce your cutoff to 0.01, or even lower... For me, in the  
end, it depends on how many mistakes can I live with. If I pick up  
2000 genes, but I expect 100 to be wrong... depending on the actual  
experiment I might go for a stricter cutoff where I may only pick 100,  
but expecting only 1 to be wrong. In other cases a higher rate of  
error is not a problem.

In the end, as Richard said, validation by PCR is necessary, not of  
every gene, but of a reasonable number across the range.

Jose


Quoting Richard Friedman <friedman at cancercenter.columbia.edu>:

> Dear Mouselly,
>
> 	A B>f 0 is a good place to start.
> It is advisable to spot check genes by PCR to validate the cutoff.
>
> Best wishes,
> Rich
>
> On Jul 18, 2007, at 10:06 AM, el mousselly antra wrote:
>
>> hi,
>>   i have a question on  limma package
>>   when we have the result of tnis pakage and we have p-vlue and the
>>  value of B .
>>   i i want to know on wich value of p we consider that genes are DE ,
>>  we take 0.05 is the threshold or what and i see that when we consider
>>  0.05 as threshold wa have a lot of genes DE.
>>   so, i want some rule (threshold)that will be raisonnable and
>>  efficient to detect gnes DE
>>   thank you
>>   sincerly
>>
>>
>>
>>
>> ---------------------------------
>>
>> 	[[alternative HTML version deleted]]
>>
>> _______________________________________________
>> 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
>
> ------------------------------------------------------------
> Richard A. Friedman, PhD
> Biomedical Informatics Shared Resource
> Lecturer
> Department of Biomedical Informatics
> Educational Coordinator
> Center for Computational Biology and Bioinformatics
> National Center for Multiscale Analysis of Genomic Networks
> Box 95, Room 130BB or P&S 1-420C
> Columbia University Medical Center
> 630 W. 168th St.
> New York, NY 10032
> (212)305-6901 (5-6901) (voice)
> friedman at cancercenter.columbia.edu
> http://cancercenter.columbia.edu/~friedman/
>
> In memoriam, John Stewart Williamson
>
> _______________________________________________
> 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
>
>



-- 
Dr. Jose I. de las Heras                      Email: J.delasHeras at ed.ac.uk
The Wellcome Trust Centre for Cell Biology    Phone: +44 (0)131 6513374
Institute for Cell & Molecular Biology        Fax:   +44 (0)131 6507360
Swann Building, Mayfield Road
University of Edinburgh
Edinburgh EH9 3JR
UK



More information about the Bioconductor mailing list