[BioC] BSgenomeForge seed file - seqnames field

Hervé Pagès hpages at fhcrc.org
Thu May 2 19:49:10 CEST 2013


Hi Kelly,

On 05/02/2013 06:44 AM, Vining, Kelly wrote:
> Hi Herve,
> Thanks for your helpful response, and for pointing me to the proper scripts. Given this, it appears that I should treat my gene feature annotation files (promoters, genes) as mseqnames objects, similarly to how I will treat the extrachromosomal scaffolds. Is there any way to include additional features that I may find of interest after I forge this initial package (say, using gff files), or am I limited such that I would have to re-forge a new reference?

A BSgenome data package is for storing the sequences of a given
genome/assembly. No annotations should go there.

In Bioconductor we try to maintain a clear separation between
packages that contain the sequences of a genome/assembly (BSgenome
data packages), and packages that contain annotations for that
genome/assembly (TxDb, FDb, OrganismDb, SNPlocs, etc... packages).

There are at least 2 reasons for this:

   1. For a given assembly there are many kinds of annotations available
      in many places on the internet, and some of them tend to be updated
      frequently. By contrast, the sequences of a given assembly
      never change. So putting only the reference sequences in a
      BSgenome package make it very stable. It almost never needs to
      be updated, except for updating a man page or when something
      changes in the way sequences are stored in the package. This is
      good with such big packages (800 Mb for the biggest ones).

   2. Storing the DNA sequences of the genes, promoters, or other genomic
      feature in a BSgenome package would duplicate a lot of DNA
      sequences that are already in the reference sequences (those
      small sequences are substrings of the reference sequences).
      This would make the BSgenome package unnecessarily bigger.
      It's better to store only the locations of those genomic features,
      not in the BSgenome package, but in a separate package, and then
      to use those locations to extract the corresponding sequences from
      the BSgenome object. This extraction can be done with getSeq()
      (defined in the BSgenome software package) and is relatively
      cheap. This approach gives you a lot more flexibility than having
      those sequences pre-extracted and bundled with the BSgenome data
      package.

      If your annotations are in GFF files, you don't really need to
      put them in a package: you can load them directly in GRanges
      objects with import() (defined in the rtracklayer package),
      or if you are particularly interested in the exon/transcript
      structure of your genes, you could use makeTranscriptDbFromGFF()
      to load a GFF file containing genes, mRNAs, exons, and CDSs into
      a TranscriptDb. See ?makeTranscriptDbFromGFF in the
      GenomicFeatures package for more information. Then it's easy to
      extract the locations of transcripts, exons, or CDSs as a GRanges
      or GRangesList object from this TranscriptDb object. See
      ?transcripts and ?exonsBy. Finally once you have the locations
      in a GRanges object, you can use getSeq() to extract the
      corresponding sequences from the BSgenome object.

      Other more specialized functions are promoters() and
      getPromoterSeq() for extracting the promoter locations and
      their sequences, respectively. Also extractTranscriptsFromGenome()
      for extracting the full transcriptome from a BSgenome package.

HTH,
H.

PS: The upstream sequences that you see in some of our BSgenome data
packages are a relic of the past and will be removed soon.

>
> Thanks much,
> --Kelly V.
>
> -----Original Message-----
> From: Hervé Pagès [mailto:hpages at fhcrc.org]
> Sent: Tuesday, April 30, 2013 11:12 PM
> To: Kelly V [guest]
> Cc: bioconductor at r-project.org; Vining, Kelly; MEDIPS Maintainer
> Subject: Re: [BioC] BSgenomeForge seed file - seqnames field
>
> Hi Kelly,
>
> On 04/30/2013 03:52 PM, Kelly V [guest] wrote:
>>
>> I'm preparing a custom reference genome for use with the MEDIPS package. I see that one field of the seed file, which is apparently not optional, is the 'seqnames' field. The example given in the documentation is this:
>>
>> paste("chr", c(1:20, "X", "M", "Un", paste(c(1:20, "X", "Un"),
>> "_random", sep="")), sep="")
>>
>> I have two simple questions about this.
>>
>> 1. Does R match this information with the source sequence file? For example, if I have a single fasta file with fasta headers chr_01...chr_20, must the seqnames entries exactly match those headers?
>
> No. You need to provide 1 FASTA file per single sequence, that is, 1 file per name you put in the 'seqnames' field. That means that each file is expected to contain only 1 sequence. What's in the FASTA header of each file is not important. What's important is that the name of each file be of the form <prefix><seqname><suffix>, where <seqname> is the name of the sequence as it appears in the 'seqnames' field, and <prefix> and <suffix> are a prefix and a suffix (eventually empty) that are the same for all the files.
>
> If what you have is a big FASTA file containing all the chromosome sequences, then you first need to split it into 1 file per chromosome.
> This is easy to do in R. For example, here is the script I used to split the big bosTau6.fa file provided by UCSC for the bosTau6 genome:
>
>     library(Biostrings)
>     bosTau6 <- readDNAStringSet("bosTau6.fa")
>
>     ### Partitioning:
>     is_chrUn <- grepl("^chrUn", names(bosTau6))
>     is_chrom <- !is_chrUn
>
>     ### Send each chromosome to a FASTA file.
>     seqnames <- paste("chr", c(1:29, "X", "M"), sep="")
>     stopifnot(setequal(seqnames, names(bosTau6)[is_chrom]))
>
>     for (seqname in seqnames) {
>       seq <- bosTau6[match(seqname, names(bosTau6))]
>       filename <- paste(seqname, ".fa", sep="")
>       cat("writing ", filename, "\n", sep="")
>       writeXStringSet(seq, file=filename, width=50L)
>     }
>
>     ### Send the 3286 chrUn_* sequences to 1 FASTA file.
>     chrUn_mseq <- bosTau6[is_chrUn]
>     writeXStringSet(chrUn_mseq, file="chrUn.fa", width=50L)
>
> The input is the bosTau6.fa file containing 3317 FASTA records:
> 31 records for the chromosomes, and 3286 for the chrUn_* sequences.
> The script produces 32 FASTA files: 1 per chromosome (chr1.fa, chr2.fa, chr3.fa, ..., chr29.fa, chrX.fa, chrM.fa), and the chrUn.fa file (containing the 3286 chrUn_* sequences).
>
> Note that you can find this script in the BSgenome package, and display its source with:
>
>     > splitbigfasta_R <- system.file("extdata",
>                                      "GentlemanLab",
>                                      "BSgenome.Btaurus.UCSC.bosTau6-tools",
>                                      "splitbigfasta.R",
>                                      package="BSgenome")
>
>     > cat(readLines(splitbigfasta_R), sep="\n")
>
> It should not be too hard to adapt this script to your own needs.
>
>>
>> 2. Revealing the reason for my first question:In my genome fasta file, I have 1427 extrachromosomal scaffolds, but they are not all sequentially numbered, so that I have scaffold_1..scaffold_3681. Do I need to use a regular expression in my seqnames field to tell R to look for scaffold_ followed by 1-4 digits?
>
> No, not in the 'seqnames' field, because those 1427 extrachromosomal scaffolds should not go there. They would need to go in the 'mseqnames'
> field.
>
> Unlike the 'seqnames' field where you enumerate objects that can only contain 1 sequence, in the 'mseqnames' field you can enumerate objects that contain more than 1 sequence. More precisely, in the final BSgenome data package, each entry in the 'seqnames' field will correspond to a DNAString object (the DNAString container can hold
> 1 sequence only), and each entry in the 'mseqnames' field will correspond to a DNAStringSet object (the DNAStringSet container can hold multiple sequences).
>
> So typically, all the extrachromosomal scaffolds would go in one DNAStringSet object in the final BSgenome package (this is for example what is done in the BSgenome.Drerio.UCSC.danRer7 package).
> To achieve this, you need to put 1 entry in the 'mseqnames' field (e.g. "scaffolds"), and to put the 1427 extrachromosomal scaffolds in one FASTA file named accordingly to that entry (e.g. scaffolds.fa).
>
> In the above script, replace:
>
>     is_chrUn <- grepl("^chrUn", names(bosTau6))
>
> with:
>
>     is_scaffold <- grepl("^scaffold", names(<your_genome>))
>
> then every occurrence of 'is_chrUn' with 'is_scaffold', and finally those 3 lines:
>
>     ### Send the 3286 chrUn_* sequences to 1 FASTA file.
>     chrUn_mseq <- bosTau6[is_chrUn]
>     writeXStringSet(chrUn_mseq, file="chrUn.fa", width=50L)
>
> with:
>
>     ### Send the 1427 scaffold_* sequences to 1 FASTA file.
>     scaffolds_mseq <- <your_genome>[is_scaffold]
>     writeXStringSet(scaffolds_mseq, file="scaffolds.fa", width=50L)
>
> and that should take care of sending all the scaffold sequences to the scaffolds.fa file.
>
> Let me know if you need further assistance with this.
>
> Cheers,
> H.
>
>>
>> Thanks for any help,
>> --Kelly V.
>>
>>    -- output of sessionInfo():
>>
>> R version 3.0.0 (2013-04-03)
>> Platform: i386-w64-mingw32/i386 (32-bit)
>>
>> locale:
>> [1] LC_COLLATE=English_United States.1252 [2] LC_CTYPE=English_United
>> States.1252 [3] LC_MONETARY=English_United States.1252 [4]
>> LC_NUMERIC=C [5] LC_TIME=English_United States.1252
>>
>> attached base packages:
>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>
>> --
>> Sent via the guest posting facility at bioconductor.org.
>>
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives:
>> http://news.gmane.org/gmane.science.biology.informatics.conductor
>>
>
> --
> 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 fhcrc.org
> Phone:  (206) 667-5791
> Fax:    (206) 667-1319
>

-- 
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 fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319



More information about the Bioconductor mailing list