[BioC] select top genes based on p-value in limma

James MacDonald jmacdon at med.umich.edu
Thu Mar 4 20:59:07 MET 2004


topTable normally sorts genes based on the log odds of differential
expression (the 'B' statistic). You can sort by any criterion using the
sort.by= call to topTable e.g.,

topTable(fit3, sort.by="P") #sorts by the p-value

from ?topTable

sort.by: statistic to rank genes by.  Possibilities are '"M"', '"A"',
          '"T"', '"P"' or '"B"'.

Another way to get the top genes by a p-value is to do 

tt <- topTable(fit, coef=1, number=200, sort.by="P", genelist=genelist,
adjust="fdr")
tt <- tt[tt[,4]<0.05,]

then tt only contains genes with a p-value < 0.05

HTH,

Jim



James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623

>>> <Arne.Muller at aventis.com> 03/04/04 12:28PM >>>
Hello,

I'm converting the limma fit into a data frame:

...
> fit3 <- eBayes(fit2)
> d <- data.frame(fit3$"p.value", fit3$'lods')
> colnames(d) <- c('pv', 'fc')
> d[1:5,]
           pv        fc
1 0.002538643 -1.470819
2 0.802832281 -6.320708
3 0.545533328 -6.155623
4 0.688787086 -6.267272
5 0.312317306 -5.799080
> 

then suppose gnames is a vector of gene names, I select all genes with
a
p-value from the eBayes calculation <= 0.05:

> mygenes <- gnames[which(d[,'pv'] <= 0.05)]

Actually I'm still exploring the limma package, and I'm not sure what
topTable actually does - it doesn' seem to return the genes by ordered
p-value.

	regards,

	Arne

--
Arne Muller, Ph.D.
Toxicogenomics, Aventis Pharma
arne dot muller domain=aventis com

> -----Original Message-----
> From: bioconductor-bounces at stat.math.ethz.ch 
> [mailto:bioconductor-bounces at stat.math.ethz.ch]On Behalf Of p hu
> Sent: 04 March 2004 01:02
> To: bioconductor at stat.math.ethz.ch 
> Subject: [BioC] select top genes based on p-value in limma
> 
> 
> Hi all,
>  
> For example, I used 
>  
> clas<-classifyTests(fit,p.value=0.05)
> mycount<-vennCounts(clas, include="both")
>  
> and found there are 99 differentially expressed genes for my 
> first comparsion.
>  
> then I do:
>  
> toptable1<-topTable(fit,coef=1,number=99,genelist=genelist,adj
> ust="fdr")
>  
> this is part of the results
>           Name          M         t      P.Value           B
> 6820    H25306  2.9578622 21.779101 6.712472e-22 39.07274570
> 6222    H25611  3.8097340 20.616434 3.472310e-21 37.73887184
> 4394    H12333  2.9868665 13.285336 1.033728e-13 25.92910739
> 2269    R31747  3.9112632 12.339976 1.164282e-12 23.86703473
> 9171    R31507  3.7780976 11.834938 4.149866e-12 22.70578573
> 11306 AA043477  1.6451826  9.087753 2.067361e-08 15.64724863
> 596     H83378  1.1806774  7.498940 3.972530e-06 11.03899927
> 9544    H42051  1.5306360  7.202912 9.726446e-06 10.14827317
> 11320 AA054300  0.9058530  6.899348 2.492575e-05  9.22757718
> 10132 AA135957  0.8268765  6.552645 7.535515e-05  8.16932543
> 17941 AA149043  1.2592684  6.404648 1.149187e-04  7.71612881
> 13461 AA211825  0.7310730  6.082586 3.242677e-04  6.72862309
> ............................................................
> 16930   W32999 -0.3562904 -2.861632 3.849730e-01 -2.36371124
> 17667   W67427  0.4080262  2.859229 3.862250e-01 -2.36921005
> 7769    H53894 -0.3463782 -2.856329 3.879989e-01 -2.37584026
> 9751    W92088 -0.3139404 -2.854464 3.887059e-01 -2.38010239
> 5067    H57545  0.8197179  2.851609 3.902150e-01 -2.38662381
> 9468    R27989  0.3628438  2.848099 3.902150e-01 -2.39463557
>  
> As I can see here, the last gene has very high p-value 
> although it is called DE gene.
>  
> So I am wondering how I can select genes based on a cut off 
> p-value rather than a number that indicates how many genes I 
> want to pick???
>  
> Thanks
> 
> 
> 
> 
> ---------------------------------
> Post your free ad now! Yahoo! Canada Personals
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch 
> https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor 
>

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch 
https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor



More information about the Bioconductor mailing list