[BioC] new topGO results using GO.db very different from old ones using GO

Adrian Alexa adrian.alexa at gmail.com
Fri May 2 17:07:08 CEST 2008


Hi Joern,

seems like a bug in topGO. The GO IDs are the same in both cases, but
the names are wrong in the second table. GO:0007610 is not
"reproduction" but "behavior". At the first sight looks like a bug in
the the GenTable function. but I need to look closely. There will be
an update soon.

Thanks for the report and I apologies for the bug,
Adrian



On Fri, May 2, 2008 at 2:48 PM, Joern Toedling <toedling at ebi.ac.uk> wrote:
> Dear all,
>  I would appreciate any suggestion on the following issue. I have noticed a
> major inconsistency between new and older topGO results. For the older ones,
> topGO used the "GO" package, while it uses "GO.db" for the new results I
> can't figure out whether it is a problem with topGO only or whether there
> are some serious inconsistencies between GO and GO.db
>
>  Here is the source code I used:
>
>  library("topGO")
>
>  ## load list of genes of interest
>
>  load("brainOnlyGenes.RData")
>
>  ## load genereal gene-to-GO mapping and universe of genes to use in
> analysis:
>
>  load("mm9gene2GO.RData")
>
>  load("arrayGenesWithGO.RData")
>
>  ## then the function to call topGO and to return a nice result table:
>
>  sigGOTable <- function(selGenes, GOgenes=arrayGenesWithGO,
>  gene2GO=mm9.gene2GO[arrayGenesWithGO], ontology="BP", maxP=0.001)
>
>  {
>
>   inGenes <- factor(as.integer(GOgenes %in% selGenes))
>
>   names(inGenes) <- GOgenes
>
>   GOdata <- new("topGOdata", ontology=ontology, allGenes=inGenes,
>                annot=annFUN.gene2GO, gene2GO=gene2GO)
>
>   myTestStat <- new("elimCount", testStatistic=GOFisherTest,
>                    name="Fisher test", cutOff=maxP)
>
>   mySigGroups <- getSigGroups(GOdata, myTestStat)
>
>   sTab <- GenTable(GOdata, mySigGroups, topNodes=length(usedGO(GOdata)))
>
>   names(sTab)[length(sTab)] <- "p.value"
>
>   return(subset(sTab, as.numeric(p.value) < maxP))
>
>  }#
>
>  ## call it:
>
>  (brainRes <- sigGOTable(brainOnlyGenes))
>
>  # with  topGO_1.4.0  using GO_2.0.1
>
>  # this is:
>
>  #         GO.ID                  Term Annotated Significant Expected
> p.value
>  #  1 GO:0007268 synaptic transmission       136          44    24.46
> 3.0e-05
>  #  2 GO:0007610              behavior       180          54    32.38
> 4.4e-05
>  #  3 GO:0007409          axonogenesis       119          38    21.41
> 0.00014
>  #  4 GO:0006887            exocytosis        40          17     7.20
> 0.00026
>  #  5 GO:0007420     brain development       136          40    24.46
> 0.00066
>
>
>  # which kind of make sense if it somehow to annotate a list of interesting
> genes when investigating brain cells
>
>  ## now unfortunately using all the same gene list, universe and gene-to-GO
> mapping, and the same function as above
>
>  ##  with topGO_1.9.0  using GO.db_2.2.0, the result is:
>
>  #        GO.ID                                Term Annotated Significant
> Expected  p.value
>  # 1 GO:0007268    mitochondrial genome maintenance       137          44
> 24.65  3.7e-05
>  # 2 GO:0007610                        reproduction       180          54
> 32.39  4.4e-05
>  # 3 GO:0007409          single strand break repair       119          38
> 21.41  0.00014
>  # 4 GO:0006887     regulation of DNA recombination        40          17
> 7.20  0.00026
>  # 5 GO:0007420 regulation of mitotic recombination       136          40
> 24.47  0.00066
>
>
>  # which is obviously very, very different
>
>
>  Does anyone have an educated guess what is going on? Could it be a bug a in
> topGO? Or is the information in GO.db really different from the one in GO,
> and in that case which one is the right one?
>
>  Best regards,
>  Joern
>
>  _______________________________________________
>  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
>



More information about the Bioconductor mailing list