[Bioc-devel] Known Genes replaced by GENCODE genes at UCSC

Hervé Pagès hpages at fredhutch.org
Mon Jan 11 22:12:28 CET 2016


Hi Robert and others,

I looked at this and the new situation doesn't seem as disruptive as
it sounds. The bulk of the data for both tracks (i.e. the "UCSC Genes"
track for hg19 and the "GENCODE v22" track for hg38) is stored in the
knownGene table.

The hg19.knownGene table is described here:

 
https://genome.ucsc.edu/cgi-bin/hgTables?db=hg19&hgta_group=genes&hgta_track=knownGene&hgta_table=knownGene&hgta_doSchema=describe+table+schema

The hg38.knownGene table is described here:

 
https://genome.ucsc.edu/cgi-bin/hgTables?db=hg38&hgta_group=genes&hgta_track=knownGene&hgta_table=knownGene&hgta_doSchema=describe+table+schema

The 2 pages are very similar. In particular both tables are connected
to the knownToLocusLink table where Entrez Gene IDs are stored.

So from a makeTxDbFromUCSC() point of view everything looks the same
except the name of the track: "UCSC Genes" for hg19 and "GENCODE v22"
for hg38. That means it shouldn't be hard to tweak makeTxDbFromUCSC()
to support:

     txdb <- makeTxDbFromUCSC("hg38", "knownGene")

The returned 'txdb' will contain data from the "GENCODE v22" track
and with transcripts mapped to Entrez Gene IDs.

I'll work on this and will also investigate makeTxDbFromGRanges's
failure on AnnotationHub's GFF files from GENCODE.

H.


On 01/11/2016 06:29 AM, Robert Castelo wrote:
> hi,
>
> if i'm interpreting this correctly, the news archive of the UCSC Genome
> Browser accessible here:
>
>   http://genome.ucsc.edu/goldenPath/newsarch.html
>
> announced on June 29th, 2015, that they are discontinuing the generation
> of UCSC Known Genes annotations for human, and provide the Gencode
> annotations as default replacement.
>
> the BioC site provides as default gene annotations for human the UCSC
> Known Genes track and currently does not provide the Gencode annotations.
>
> the GenomicFeatures package allows one to build such an annotation
> package. unfortunately the current "supported" UCSC tables that can be
> easily used via 'makeTxDbPackageFromUCSC()' reports up to Gencode
> version V17:
>
> library(GenomicFeatures)
>
> xx <- supportedUCSCtables()
> xx[grep("GENCODE Genes", xx$track), ]
>                                               track subtrack
> wgEncodeGencodeBasicV17          GENCODE Genes V17     <NA>
> wgEncodeGencodeCompV17           GENCODE Genes V17     <NA>
> wgEncodeGencodePseudoGeneV17     GENCODE Genes V17     <NA>
> wgEncodeGencode2wayConsPseudoV17 GENCODE Genes V17     <NA>
> wgEncodeGencodePolyaV17          GENCODE Genes V17     <NA>
> wgEncodeGencodeBasicV14          GENCODE Genes V14     <NA>
> wgEncodeGencodeCompV14           GENCODE Genes V14     <NA>
> wgEncodeGencodePseudoGeneV14     GENCODE Genes V14     <NA>
> wgEncodeGencode2wayConsPseudoV14 GENCODE Genes V14     <NA>
> wgEncodeGencodePolyaV14          GENCODE Genes V14     <NA>
> wgEncodeGencodeBasicV7            GENCODE Genes V7     <NA>
> wgEncodeGencodeCompV7             GENCODE Genes V7     <NA>
> wgEncodeGencodePseudoGeneV7       GENCODE Genes V7     <NA>
> wgEncodeGencode2wayConsPseudoV7   GENCODE Genes V7     <NA>
> wgEncodeGencodePolyaV7            GENCODE Genes V7     <NA>
>
> which is about 2 years old. current Gencode gene annotations are V24 and
> at least V22 was available at:
>
> http://hgdownload.soe.ucsc.edu/goldenPath/hg38/database
>
> before the last BioC release.
>
> according to a recent announcement at the BioC support site:
>
> https://support.bioconductor.org/p/71574
>
> AnnotationHub seems to be now the proper way to import the most recent
> Gencode annotations into BioC. however, at least in my hands, making the
> corresponding TxDb object produces an error; see the following example:
>
> library(AnnotationHub)
>
> ah <- AnnotationHub()
> human_gff <- query(ah, c("Gencode", "gff", "human"))
>
> gencodeV23basicGFF <- ah[["AH49556"]]
> metadata <- data.frame(name=c("Data source", "Genome", "Organism",
>                                "Resource URL", "Full dataset"),
>                         value=c(ah["AH49556"]$dataprovider,
> ah["AH49556"]$genome,
>                                 ah["AH49556"]$species,
> ah["AH49556"]$sourceurl, "no"))
> txdb <- makeTxDbFromGRanges(gencodeV23basicGFF, metadata=metadata)
> Error in .merge_transcript_parts(transcripts) :
>    The following transcripts have multiple parts that cannot be merged
> because of incompatible seqnames: ENST00000244174.9,
>    ENST00000262640.10, ENST00000286448.10, ENST00000302805.6,
> ENST00000313871.7, ENST00000326153.8, ENST00000331035.8,
>    ENST00000334060.7, ENST00000334651.9, ENST00000355432.7,
> ENST00000355805.6, ENST00000359512.7, ENST00000369423.6,
>    ENST00000381180.7, ENST00000381184.5, ENST00000381187.7,
> ENST00000381192.7, ENST00000381218.7, ENST00000381222.6,
>    ENST00000381223.8, ENST00000381229.8, ENST00000381233.7,
> ENST00000381241.7, ENST00000381261.7, ENST00000381297.8,
>    ENST00000381317.7, ENST00000381333.8, ENST00000381401.9,
> ENST00000381469.6, ENST00000381500.5, ENST00000381509.7,
>    ENST00000381524.7, ENST00000381529.7, ENST00000381566.5,
> ENST00000381567.7, ENST00000381575.5, ENST00000381578.5,
>    ENST00000381657.6, ENST00000381663.7, ENST00000390665.7,
> ENST00000391707.6, ENST00000399012.5, ENST00000399966.8,
>    ENST00000400841.6, ENST00000411342.5, ENST00000412936
>
>
> on top of this, even if it would work, these annotations are anchored at
> Ensembl Gene identifiers while the gene-centric annotations at
> org.Hs.eg.db are anchored at Entrez Gene identifiers. this means that
> more code would have to be involved to add the corresponding Entrez IDs
> (resolving multiplicities, etc.) and produce a TxDb package that can be
> used across many of the typical BioC pipelines.
>
> since human gene annotations are at the core of many BioC pipelines, i'd
> like to suggest for the forthcoming release cycles, that the BioC core
> team packages Gencode annotations anchored at Entrez IDs, at least what
> is called the "basic set", similarly to what is done with
> TxDb.Hsapiens.UCSC.knownGene to have an easy starting point for the
> analysis of human data.
>
>
> cheers,
>
> robert.
>
> _______________________________________________
> Bioc-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/bioc-devel

-- 
Hervé Pagès

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fredhutch.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319



More information about the Bioc-devel mailing list