Skip to content

[Bioc-devel] glitches with releaseName() and seqlevelsStyle()

4 messages · Robert Castelo, Hervé Pagès

Original
Robert Castelo
# 
hi,

i've found the following glitches with 'releaseName()' and 
'seqlevelsStyle()', i guess due to recent changes in BSgenome and 
GenomeInfoDb, which is why i'm cc'ing Herv? ;)

===========BioC release 3.11 everything OK=============
library(BSgenome.Hsapiens.UCSC.hg19)
library(TxDb.Hsapiens.UCSC.hg19.knownGene)

txdb <- TxDb.Hsapiens.UCSC.hg19.knownGene

releaseName(Hsapiens)
[1] "Genome Reference Consortium GRCh37.p13"
releaseName(as(Hsapiens, "GenomeDescription"))
[1] "Genome Reference Consortium GRCh37.p13"

seqlevelsStyle(txdb)
[1] "UCSC"
seqlevelsStyle(txdb) <- "NCBI"
seqlevelsStyle(txdb)
[1] "NCBI"??? "Ensembl"
head(seqlengths(txdb))
 ??????? 1???????? 2???????? 3???????? 4???????? 5 6
249250621 243199373 198022430 191154276 180915260 171115067
=======================================================

============BioC devel 3.12 inconsistencies============
library(BSgenome.Hsapiens.UCSC.hg19)
library(TxDb.Hsapiens.UCSC.hg19.knownGene)

txdb <- TxDb.Hsapiens.UCSC.hg19.knownGene

releaseName(Hsapiens)
Error in (function (classes, fdef, mtable)? :
 ? unable to find an inherited method for function ?releaseName? for 
signature ?"BSgenome"?
releaseName(as(Hsapiens, "GenomeDescription"))
[1] NA

seqlevelsStyle(txdb)
[1] "UCSC"
seqlevelsStyle(txdb) <- "NCBI"
seqlevelsStyle(txdb)
[1] "NCBI" "UCSC"
head(seqlengths(txdb))
 ??????? 1???????? 2???????? 3???????? 4???????? 5 6
249250621 243199373 198022430 191154276 180915260 171115067
========================================================

in the case of 'releaseName()' i interpret that the method has become 
defunct for 'BSgenome' objects, but i think it should give a sensible 
result with a 'GenomeDescription' object, so something may be going 
wrong with the coercion from 'BSgenome' to 'GenomeDescription'.

in the case of 'seqlevelsStyle()', while the change of style seems to 
happen, i don't see how the style can be simultaneously NCBI and UCSC.

below my session information for the BioC devel 3.12 version.

thanks!!

robert.

sessionInfo()
R version 4.0.0 (2020-04-24)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 18.04.4 LTS

Matrix products: default
BLAS/LAPACK: /usr/lib/x86_64-linux-gnu/libopenblasp-r0.2.20.so

locale:
 ?[1] LC_CTYPE=en_US.UTF-8?????? LC_NUMERIC=C
 ?[3] LC_TIME=en_US.UTF-8??????? LC_COLLATE=en_US.UTF-8
 ?[5] LC_MONETARY=en_US.UTF-8??? LC_MESSAGES=C
 ?[7] LC_PAPER=en_US.UTF-8?????? LC_NAME=C
 ?[9] LC_ADDRESS=C?????????????? LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats4??? parallel? stats???? graphics? grDevices utils???? datasets
[8] methods?? base

other attached packages:
 ?[1] TxDb.Hsapiens.UCSC.hg19.knownGene_3.2.2
 ?[2] GenomicFeatures_1.41.2
 ?[3] AnnotationDbi_1.51.3
 ?[4] Biobase_2.49.0
 ?[5] BSgenome.Hsapiens.UCSC.hg19_1.4.3
 ?[6] BSgenome_1.57.5
 ?[7] rtracklayer_1.49.5
 ?[8] Biostrings_2.57.2
 ?[9] XVector_0.29.3
[10] GenomicRanges_1.41.6
[11] GenomeInfoDb_1.25.10
[12] IRanges_2.23.10
[13] S4Vectors_0.27.12
[14] BiocGenerics_0.35.4
[15] BiocManager_1.30.10

loaded via a namespace (and not attached):
 ?[1] SummarizedExperiment_1.19.6 progress_1.2.2
 ?[3] tidyselect_1.1.0 purrr_0.3.4
 ?[5] lattice_0.20-41 generics_0.0.2
 ?[7] vctrs_0.3.1 BiocFileCache_1.13.1
 ?[9] blob_1.2.1 XML_3.99-0.4
[11] rlang_0.4.6 pillar_1.4.4
[13] glue_1.4.1 DBI_1.1.0
[15] rappdirs_0.3.1 BiocParallel_1.23.2
[17] bit64_0.9-7.1 dbplyr_1.4.4
[19] matrixStats_0.56.0 GenomeInfoDbData_1.2.3
[21] lifecycle_0.2.0 stringr_1.4.0
[23] zlibbioc_1.35.0 memoise_1.1.0
[25] biomaRt_2.45.2 curl_4.3
[27] Rcpp_1.0.4.6 openssl_1.4.1
[29] DelayedArray_0.15.7 bit_1.1-15.2
[31] Rsamtools_2.5.3 hms_0.5.3
[33] askpass_1.1 digest_0.6.25
[35] stringi_1.4.6 dplyr_1.0.0
[37] grid_4.0.0 tools_4.0.0
[39] bitops_1.0-6 magrittr_1.5
[41] RCurl_1.98-1.2 RSQLite_2.2.0
[43] tibble_3.0.1 crayon_1.3.4
[45] pkgconfig_2.0.3 ellipsis_0.3.1
[47] Matrix_1.2-18 prettyunits_1.1.1
[49] assertthat_0.2.1 httr_1.4.1
[51] R6_2.4.1 GenomicAlignments_1.25.3
[53] compiler_4.0.0
Hervé Pagès
# 
Hi Robert,
On 9/2/20 11:12, Robert Castelo wrote:
I'm dropping the notion of "release name" for BSgenome objects in BioC 
3.12. For backward compatibility I just restored a temporary 
releaseName() method (in BSgenome 1.57.6) that does the following:

 > releaseName(BSgenome.Hsapiens.UCSC.hg19)
[1] NA
Warning message:
   Starting with Bioconductor 3.12, BSgenome objects no longer have a
   "release name". As a consequence of this change calling releaseName()
   on a BSgenome object now always returns NA and is deprecated.
This happens when you end up with a mix of styles. In the case of hg19, 
all the sequences get renamed from UCSC to NCBI names **except** chrM:

   library(TxDb.Hsapiens.UCSC.hg19.knownGene)
   txdb <- TxDb.Hsapiens.UCSC.hg19.knownGene

   table(genome(txdb))
   # hg19
   #   93

   seqlevelsStyle(txdb) <- "NCBI"

   seqlevelsStyle(txdb)
   # [1] "NCBI" "UCSC"

   table(genome(txdb))
   # GRCh37.p13       hg19
   #         92          1

   genome(txdb)[genome(txdb) == "hg19"]
   #   chrM
   # "hg19"

chrM cannot and should not be renamed to MT because the MT sequence in 
GRCh37.p13 is NOT the same as the chrM sequence in hg19. So the new 
seqlevelsStyle() behavior fixes a long standing bug.

See previous discussion on this list about the long dance between UCSC 
and NCBI about the mitochondrian sequence in hg19/GRCh37/GRCh37.p13: 
https://stat.ethz.ch/pipermail/bioc-devel/2020-August/017086.html

Hope this helps,
H.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Robert Castelo
    

    

      
      #
    

  


  

      
Hi Herv?,

        
On 02/09/2020 21:44, Pages, Herve wrote:

            

      
      
      
    

        
ah ok, i see, this is fine, but the purpose of the 'releaseName()' 
function in the documentation is confusing, 'help(releaseName)' says:

?releaseName(x)?: Return the release name of this genome, which is
 ????????? generally made of the name of the organization who assembled
 ????????? it plus its Build version.? For example, UCSC uses ?"hg18"?
 ????????? for the version of the Human genome corresponding to the
 ????????? Build 36.1 from NCBI hence the release name for this genome
 ????????? is ?"NCBI Build 36.1"?.

but, if i'm interpreting you correctly, genomes in BSgenome.* packages 
have no release name anymore. maybe the help page of 'releaseName', 
which is the one of the 'GenomeDescription' class should say only:

'releaseName(x)': Return the release name of 'x'.

for whatever 'x' may be and whatever a release name may mean.

            

      
      
      
    

        
i see, this makes sense, i've followed the discussion about the 
mitochondrian sequence naming, but didn't connect it with this behavior.

i'd like to set the sequence style of a TxDb object to whatever the 
sequence style is of the BSgenome object, dropping the levels in the 
TxDb object that are not common to the BSgenome object (removing the 
mitochondrial chromosome in the case of hg19/GRCh37), so that the TxDb 
object has a single sequence style, i thought this would work but i get 
an error:

library(BSgenome.Hsapiens.1000genomes.hs37d5)
library(TxDb.Hsapiens.UCSC.hg19.knownGene)

txdb <- TxDb.Hsapiens.UCSC.hg19.knownGene
bsgenome <- hs37d5

seqlevelsStyle(txdb) <- seqlevelsStyle(bsgenome)
Warning message:
In .normarg_seqlevelsStyle(value) :
 ? more than one seqlevels style supplied, using the 1st one only
seqlevelsStyle(txdb)
[1] "NCBI" "UCSC"

commonChr <- intersect(seqlevels(txdb), seqlevels(bsgenome))
commonChr
 ?[1] "1"? "2"? "3"? "4"? "5"? "6"? "7"? "8"? "9"? "10" "11" "12" "13" 
"14" "15"
[16] "16" "17" "18" "19" "20" "21" "22" "X"? "Y"
txdb <- keepSeqlevels(txdb, commonChr)
seqlevelsStyle(txdb)
Error in .normarg_genome(value, seqnames(x)) :
 ? when length of supplied 'genome' vector is not 1, then it must equal
 ? the number of sequences
traceback()
9: stop(wmsg("when length of supplied 'genome' vector is not 1, ",
 ?????? "then it must equal the number of sequences"))
8: .normarg_genome(value, seqnames(x))
7: `genome<-`(`*tmp*`, value = x$user_genome)
6: `genome<-`(`*tmp*`, value = x$user_genome)
5: seqinfo(x)
4: seqinfo(x)
3: seqlevelsStyle(seqinfo(x))
2: seqlevelsStyle(txdb)
1: seqlevelsStyle(txdb)

i guess this should have worked, right?

thanks!

robert.

            

      
      
      
    

  
  


  


      

    

    
15 days later

    

      
      

        


  

        

      Hervé Pagès
    

    

      
      #
    

  


  

      
oops, sorry for letting this slip off my radar.

see below...

        
On 9/3/20 05:22, Robert Castelo wrote:

            

      
      
      
    

        
Fair point. Since releaseName() will go away at some point (not just the 
method for BSgenome objects but also other methods), I clarified the man 
page:

     releaseName(x):
       IMPORTANT NOTE: The 'releaseName()' methods for GenomeDescription
       and BSgenome objects are deprecated in BioC 3.12!

       If 'x' is a GenomeDescription object, 'releaseName(x)'
       returns the release name of this genome, which is usually made
       of the name of the organization who assembled it plus its Build
       version. For example, UCSC uses "hg18" for the version of
       the Human genome corresponding to the Build 36.1 from NCBI hence
       the release name for this genome is "NCBI Build 36.1".

       If 'x' is a BSgenome object, releaseName(x) returns NA_character_

Hopefully this is clearer now.

            

      
      
      
    

        
Yes it should. Should work now with GenomicFeatures 1.41.3.

Don't hesitate to let me know if you run into other rough edges like 
this and sorry for the inconvenience.

Cheers,
H.