[Bioc-devel] VRanges with multiple samples

Robert Castelo robert.castelo at upf.edu
Wed Jan 28 17:47:40 CET 2015


currently, the VariantFiltering package works with GRanges objects 
obtained from locateVariants() and predictCoding() to hold annotated 
variants and add further annotations. However, I'd like to use 'VRanges' 
objects which are, as far as i understand them, developed for exactly 
the purpose of storing and manipulating variants and their annotations.

from the available documentation, it seems to me that the route for this 
should coercing the 'VCF' object obtained with 'readVcf()' to 'VRanges' 
via as(vcf, "VRanges"). When the input 'VCF' object has more than one 
sample, this results in to a 'VRanges' object with the variants 
replicated per different sample and a sample indicator column.

i was thinking that as one add annotations to variants the redundancy of 
the information stored in a "multi-sample VRanges" which greatly 
increase, so I was thinking to work having a minimal 'multi-sample 
VRanges' with the sample-specific information and store the annotations 
in a separate DataFrame object with some index column that would link to 
the corresponding 'VRanges' "row".

i'd like to as you if you have thoughts, suggestions or comments about 
this redundancy issue and this approach i'm thinking about.

btw, in the presence of multiple samples, i would find more natural to 
coerce a 'VCF' object into a VRangesList object, instead of a VRanges 
with a sample indicator column.

there is in fact the 'stackSamples()' method to compress a 'VRangesList' 
into a 'VRanges' with a sample indicator column, however there is no 
coercion method:

as(vcf, "VRangesList")
Error in as(vcf, "VRangesList") :
   no method or default for coercing “CollapsedVCF” to “VRangesList”

i guess i can write some one-liner get a 'VRangesList' from a 
multi-sample VCF with one 'VRanges' element per sample, but i wonder 
whether it would not make sense to have this as an 'as(vcf, 
"VRangesList")' method.


More information about the Bioc-devel mailing list