[BioC] Rsamtools filtering bam file: RE: MEDIPS.createSet error

Martin Morgan mtmorgan at fhcrc.org
Tue Apr 1 22:49:20 CEST 2014


On 04/01/2014 12:52 PM, Vining, Kelly wrote:
> This is a follow-up question about the most efficient way to use functions in Rsamtools to filter my bam files so that only alignments to chromosomes are included (extrachromosomal scaffold alignments are excluded).

I think from your original question you are really looking to provide the names 
of the sequences in your BSgenome object as a value to the chr.select argument 
of MEDIPS.createSet, I *think* chr.select=seqnames(BSgenome) (be sure to only 
include seqnames defined in the bam file, too).

>
>>From the Rsamtools documentation, it looks like there are two ways to accomplish this:
> 1) bamFile(bamFilter())

??? I'm not sure what you're talking about here, neither bamFile nor bamFilter 
are functions defined in Rsamtools ???

Use filterBam() to create a new bam file from an old bam files. It's use is 
illustrated in an example on the help page ?filterBam.

> 2) bamFile(ScanBamParam())

ScanBamParam() is a way to selectively input data, e.g., with the 
readGAlignments or scanBam functions. Its use is illustrated for instance on p. 
2 of the vignette available in the Rsamtools package and at

http://bioconductor.org/packages/release/bioc/vignettes/Rsamtools/inst/doc/Rsamtools-Overview.pdf

MEDIPS.createSet uses a ScanBamParam internally, but as a user you do not have 
access to it.

Martin

>
> For the first method, a "destination" output file name has to be designated. I don't want to create a separate file, so it appears that the second method is what I want, as it allows control of which records are imported and doesn't output a separate file.
>
> I set up a FilterRules object like so:
>> filt <- FilterRules(list(isChr = function(x) rname(x) == "^Chr"))
>
> I also set up a "which" variable to contain the seqnames in my BSgenome:
>> which <- seqnames(BSgenome)
>
> However, I'm getting stuck trying to apply either of these to ScanBamParam. Perhaps I want ScanBamWhat instead of ScanBamParam? Can someone clue me in about the proper syntax to accomplish this simple filtering operation?
>
> Thanks for any help,
> --Kelly
>
> ________________________________________
> From: bioconductor-bounces at r-project.org [bioconductor-bounces at r-project.org] on behalf of Vining, Kelly [Kelly.Vining at oregonstate.edu]
> Sent: Tuesday, April 01, 2014 8:12 AM
> To: 'Lukas Chavez'
> Cc: Steve Lianoglou; bioconductor at r-project.org
> Subject: Re: [BioC] MEDIPS.createSet error
>
> I see, Lukas. That makes sense. Thanks for explaining it. So, MEDIPS looks in the BSgenome object for seqnames only, and doesn't look in mseqnames. It seems like a lot of users of the MEDIPS package must encounter this problem, because most genomes have sets of unassembled scaffolds that, following the BSgenome forging instructions, inevitably end up as mseq objects. It would be great if a future version of MEDIPS looked for both types of sequences in the BSgenome.
>
> I also meant to reply to Steve's question about this line from the vignette:
> BSgenome="BSgenome.Hsapiens.UCSC.hg19"
>
> In my case, with my custom genome, this would be:
> BSgenome = "BSgenome.Ptrichocarpa.Phytozome.v3"
>
> But when I do this, the literal BSgenome.Ptrichocarpa.Phytozome.v3 is assigned to the BSgenome variable as a character vector. So instead, I do this:
> BSgenome = BSgenome.Ptrichocarpa.Phytozome.v3
>
> That works. So maybe quotes are only required for genomes listed under available_genomes()?
>
> Meanwhile, for my MEDIPS analysis, I'm willing to go back to the bam files and filter out the scaffold alignments. It looks like Rsamtools has methods for this. Should I be trying to set up a FilterRules instance, or a ScanBamParam instance? Any help with the syntax for this step will be appreciated.
>
> I'm sure this is not going to work, but I'll start with something like this:
> bamFilt_Fall1 <- BamFile(filterBam(file, filter=FilterRules("scaffold")))
>
>
> Thanks much,
> --Kelly
>
> From: Lukas Chavez [mailto:lukas.chavez.mailings at googlemail.com]
> Sent: Tuesday, April 01, 2014 12:32 AM
> To: Vining, Kelly
> Cc: Steve Lianoglou; bioconductor at r-project.org
> Subject: Re: [BioC] MEDIPS.createSet error
>
> Thank you Steve for explaining in detail how to look at the content of the BSgenome object and the bam file.
>
> When MEDIPS imports the bam file, it will try to extract information for the 'scaffold_' chromosomes from the BSgenome object. The initially reported error
>
> Calculating genomic coordinates...Error in vector(length = supersize_chr[length(
> chromosomes)], mode = "character") :
>    vector size cannot be NA/NaN
> occurs because MEDIPS is trying to find the length of a chromosome given in the bam file, but does not find the chromosome in the BSgenome object.  Subsequently, MEDIPS tries to calculate genomic coordinates for the chromosome wide windows, but the length of the chromosome is NA. I understand that it will be necessary to catch this error and add a warning message in a future version of MEDIPS.
> If you want to keep your hits on the 'scaffold_' chromosomes in the bam file, and to include all scaffolds in your further analysis, you will have to recreate your BSgenome object treating each scaffold as an individual chromosome just as the other assembled chromosomes.
> Lukas
>
>
>
> On Mon, Mar 31, 2014 at 3:05 PM, Vining, Kelly <Kelly.Vining at oregonstate.edu<mailto:Kelly.Vining at oregonstate.edu>> wrote:
> Thanks, Steve, for your advice - appreciate you working with a package that isn't familiar to you.
> Thanks also for catching that I didn't have Rsamtools installed.
>
> I think this output that you suggested I generate gives an idea about what might be going on:
>
>
>> si.bsg
> Seqinfo of length 19
> seqnames seqlengths isCircular genome
> Chr01      50495391      FALSE    3.0
> Chr02      25263035      FALSE    3.0
> Chr03      21816808      FALSE    3.0
> Chr04      24267051      FALSE    3.0
> Chr05      25890704      FALSE    3.0
> ...             ...        ...    ...
> Chr15      15278577      FALSE    3.0
> Chr16      14494361      FALSE    3.0
> Chr17      16080358      FALSE    3.0
> Chr18      16958300      FALSE    3.0
> Chr19      15942145      FALSE    3.0
>> si.bam
> Seqinfo of length 1446
> seqnames      seqlengths isCircular genome
> Chr01           50495391       <NA>   <NA>
> Chr02           25263035       <NA>   <NA>
> Chr03           21816808       <NA>   <NA>
> Chr04           24267051       <NA>   <NA>
> Chr05           25890704       <NA>   <NA>
> ...                  ...        ...    ...
> scaffold_3584       3654       <NA>   <NA>
> scaffold_3595       2008       <NA>   <NA>
> scaffold_3648       2796       <NA>   <NA>
> scaffold_3664       3053       <NA>   <NA>
> scaffold_3681       1022       <NA>   <NA>
>
>
> When I created the reference genome structure, the scaffolds were all entered in a single, multiple seq object, following instructions. But in my bam files, the scaffold hits are as shown above. So that could be one problem, but I don't know how to solve that other than filtering out all of the alignments to the scaffolds from the bam files. I really would like to retain the scaffold alignments and not lose all of that precious data.
>
> The other potential issue is the "NA" entries in the "isCircular" and "genome" columns. I'm not sure whether those would be a problem.
>
> Continued thanks,
> --Kelly
>
> ________________________________________
> From: Steve Lianoglou [lianoglou.steve at gene.com<mailto:lianoglou.steve at gene.com>]
> Sent: Monday, March 31, 2014 1:15 PM
> To: Vining, Kelly
> Cc: Lukas Chavez; bioconductor at r-project.org<mailto:bioconductor at r-project.org>
> Subject: Re: [BioC] MEDIPS.createSet error
> Hi,
>
> Caveat being that I've never used MEDIPS, and I'm just going along with
> the vignette.
>
> So, comments inline:
>
>
> On 31 Mar 2014, at 13:09, Vining, Kelly wrote:
>
>> Hi,
>> Thanks for the advice thus far! To confirm what is in my BSgenome
>> variable, I did this:
>>
>>> class(BSgenome)
>> [1] "BSgenome"
>> attr(,"package")
>> [1] "BSgenome"
>>> names(BSgenome)
>> [1] "Chr01" "Chr02" "Chr03" "Chr04" "Chr05" "Chr06" "Chr07" "Chr08"
>> "Chr09"
>> [10] "Chr10" "Chr11" "Chr12" "Chr13" "Chr14" "Chr15" "Chr16" "Chr17"
>> "Chr18"
>> [19] "Chr19" "scaf"
>>
>> And then:
>>> print(BSgenome)
>> Black cottonwood genome
>> |
>> | organism: Populus trichocarpa (Black cottonwood)
>> | provider: Phytozome (JGI)
>> | provider version: 3.0
>> | release date: January 2010
>> | release name: Populus trichocarpa v3.0
>> |
>> | single sequences (see '?seqnames'):
>> |   Chr01  Chr02  Chr03  Chr04  Chr05  Chr06  Chr07  Chr08  Chr09
>> Chr10  Chr11
>> |   Chr12  Chr13  Chr14  Chr15  Chr16  Chr17  Chr18  Chr19
>> |
>> | multiple sequences (see '?mseqnames'):
>> |   scaf
>> |
>> | (use the '$' or '[[' operator to access a given sequence)
>>
>>
>> So that looks ok. Interestingly, when I followed the vignette and did
>> the equivalent of
>> BSgenome="BSgenome.Hsapiens.UCSC.hg19"
>
> The vignette suggests that BSgenome should be the package name of the
> BSgenome package to open, so yours should be something like
> "BSgenome.Ptrichocarpa.XXX.YY" or something -- I guess you built this
> packge by yourself, or something, so you'd know its name ...
>
>> That didn't work for me. It only worked without quotes. If I included
>> quotes, it just assigned a character vector to that variable.
>
> Sorry, what exactly didn't work for you? Can you show me the code that
> failed?
>
>> Then, following your advice:
>>
>>> si.bsg <- seqinfo(BSgenome.Ptrichocarpa.Phytozome.v3)
>>> si.bam <- seqinfo(BamFile("FallBud_brep1_aln_sorted.bam"))
>> Error in seqinfo(BamFile("FallBud_brep1_aln_sorted.bam")) :
>> error in evaluating the argument 'x' in selecting a method for
>> function 'seqinfo': Error: could not find function "BamFile"
>
> The BamFile function is defined in the Rsamtools package, you need to
> load that first (the first line of code I suggested you run was to load
> the Rsamtools package). Load it first, then redo the
> seqinfo(BamFile(...)) stuff.
>
> -steve
>
>
>          [[alternative HTML version deleted]]
>
> _______________________________________________
> 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
>
> _______________________________________________
> 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
>


-- 
Computational Biology / Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N.
PO Box 19024 Seattle, WA 98109

Location: Arnold Building M1 B861
Phone: (206) 667-2793



More information about the Bioconductor mailing list