[BioC] Xps package : errors and RMA difference with Partek GS

[cstrato]

Dear Arnaud,

I have just uploaded a new version xps_1.2.10 (and xps_1.3.12), which 
corrects the problem with fdr adjustment. Furthermore, I have added 
option "BY" (see ?uniTest) and crosschecked the results against 
R-function "p.adjust", thus the results should now be ok.

Thank you for reporting this problem.
Best regards
Christian


arnaud Le Cavorzin wrote:
> Dear Christian,
>
> Thank you very much for your answer and for your reactivity.
>
> We have performed RMA normalization using exonlevel=c(16316,8192,8192) 
> that is to say using all probes for the background correction and only 
> metacore probes for the quantile normalization and summarization, so 
> the same options using by Partek ("core" in Partek corresponding to 
> "metacore" in xps, like you have suggested it).
>
> We get the same number of probeset with Partek and xps package, but 
> the results are differents for the two softwares.
> Even if it was better, we found 3539 genes with a p-value<0.05 with 
> xps, and 3337 genes with p-value<0.05 with Partek, and the results 
> remain different for the two softwares. We don't obtain the same 
> p-value, in particular because we don't obtain the same means.
>
> I have also imported the data.rma from xps to Partek, and performed 
> the t test with Partek : the results are the same than performing 
> unifilter with xps, we obtain the same p value than with xps and the 
> same means. So they are still different than the results using Partek 
> only.
> (Confirm that there is a difference with the probes used in Partek for 
> the RMA normalization)
>
> Another question : when I use fdr correction or no correction with 
> xps, the results are still the same. Only when I use bonferroni 
> correction the p-adjusted change. I don't understand why FDR 
> correction have no effect.
>
> /> 
> unifltr=UniFilter(unitest=c("t.test","two.sided","none",0,0.0,FALSE,0.95,TRUE),
> + foldchange=c(1,"both"),unifilter=c(0.05,"pval"))
> > 
> rma.ufr=unifilter(data.test2.rma,"tmpdt_HuextestUnifilterallmetmet",getwd(),logbase="log2",
> + unifltr,group=c("075","075","TEM","TEM"),verbose=FALSE)
> /
> /> 
> unifltr=UniFilter(unitest=c("t.test","two.sided","fdr",0,0.0,FALSE,0.95,TRUE),
> + foldchange=c(1,"both"),unifilter=c(0.05,"pval"))
> > 
> rma.ufr=unifilter(data.test2.rma,"tmpdt_HuextestUnifilterallmetmet",getwd(),logbase="log2",
> + unifltr,group=c("075","075","TEM","TEM"),verbose=FALSE)/
>
>
> Thanks
> Best regards,
>
> Arnaud
>