[Bioc-devel] PANTHER.db GOSLIM_ID hosed?

Dear Steve,

thanks a lot for pointing out the error! The problem Martin described is
exactly the problem addressed in PANTHER.db version 1.02. However the
version available on bioconductor is not v1.02.
The method type2col in PANTHER.db 1.02 should look like this:

type2col <- function(x){

  column_names <-
c("family_id","goslim_id","ontology","uniprot_id","species","family_term","subfamily_term","class_id","class_term","go_id","go_term","component_go_id","component_term","evidence","evidence_type","confidence_code","entrez_id")
  names(column_names) <- .cols(x)
  column_names
}


What's puzzling to me is that there are unit tests included in the loading
routine, so checking the version currently available on bioconductor should
have choked executing PANTHER.db:::.test()
I remember when I first uploaded the new version to Jim, I have send an
old version by accident and then later on the correct version. This might
have caused some confusion.
Can somebody send me another aspera invite, then we can make sure that you
have got the right version?

Many thanks and apologies if this caused any troubles,
Julius



On 20/11/2015 09:46, Morgan, Martin wrote:

From the head of


PANTHER.db:::.keys

function (x, keytype)
{
    t2t <- .type2table(x)
    t2c <- .type2col(x)

I wonder if many of the columns / keytypes are mislabeled, e.g.,


PANTHER.db:::.type2col(PANTHER.db)

         CLASS_ID        CLASS_TERM      COMPONENT_ID    COMPONENT_TERM
      "family_id"       "goslim_id"        "ontology"      "uniprot_id"
  CONFIDENCE_CODE            ENTREZ          EVIDENCE     EVIDENCE_TYPE
        "species"     "family_term"  "subfamily_term"        "class_id"
        FAMILY_ID       FAMILY_TERM         GOSLIM_ID       GOSLIM_TERM
     "class_term"           "go_id"         "go_term" "component_go_id"
       PATHWAY_ID      PATHWAY_TERM           SPECIES    SUBFAMILY_TERM
 "component_term"        "evidence"   "evidence_type" "confidence_code"
          UNIPROT
      "entrez_id"

probably because columns() is used in .type2* but was changed at some point to return sorted values


showMethods(PANTHER.db:::.type2col, includeDefs=TRUE)

Function: .type2col (package PANTHER.db)
x="PANTHER.db"
function (x)
{
    cls <- c("family_id", "goslim_id", "ontology", "uniprot_id",
        "species", "family_term", "subfamily_term", "class_id",
        "class_term", "go_id", "go_term", "component_go_id",
        "component_term", "evidence", "evidence_type", "confidence_code",
        "entrez_id")
    names(cls) <- columns(x)
    cls
}


columns

nonstandardGenericFunction for "columns" defined from package "AnnotationDbi"

function (x)
{
    value <- standardGeneric("columns")
    sort(value)
}


selectMethod("columns", "PANTHER.db")

Method Definition:

function (x)
{
    c("FAMILY_ID", "GOSLIM_ID", "GOSLIM_TERM", "UNIPROT", "SPECIES",
        "FAMILY_TERM", "SUBFAMILY_TERM", "CLASS_ID", "CLASS_TERM",
        "PATHWAY_ID", "PATHWAY_TERM", "COMPONENT_ID", "COMPONENT_TERM",
        "EVIDENCE", "EVIDENCE_TYPE", "CONFIDENCE_CODE", "ENTREZ")
}

Probably a better design is to define a package-global constant

.TYPE2COLUMN = c(FAMILY_ID="family_id", GOSLIM_ID="goslim_id", ...)

and implement columns,PANTHER.db-method as names(.TYPE2COLUMN)

Martin Morgan