[Bioc-devel] InteractionSet for structural variants

Daniel Cameron d@n|e|@|@c@meron @end|ng |rom gm@||@com
Tue May 21 14:57:34 CEST 2019

>And actually, is there any package that would load a SV VCF by lumpy or
delly and build that object?

It sounds like you're looking for my StructuralVariantAnnotation package.
It's designed to load SV calls in any of various representations supported
by VCF (direct REF/ALT sequence differences, symbolic alleles, breakend
notation, and single breakend notation) and convert to a standard format
usable for downstream analysis (the initial release uses a breakend-level
GRanges object but I plan to also support different representation in the
future). I've tested against gridss, manta, delly, lumpy, crest, tigra,
pindel, gasvpro, clever, (and breakdancer/socrates in conjunction with
companion scripts to convert to VCF) although I've just noticed that delly
has changed their output format from one VCF non-complaint format to a
different VCF non-compliant format (delly is only writing one of the two
required BND records) so the package won't load results from the latest
delly version.

What's your use case? I've found it useful for a wide range of downstream
analysis including sv benchmarking (matching SV calls between
callers/between truth sets), gene disruption/fusion detection, somatic
filtering, breakpoint graph traversal, consistency checking between
breakpoint and CNV calls, and others.

> them in one object, just lump them into a List with "translocation"
(GInteractions), "cnv" (GRanges) and "inversion" (another
GRanges) elements, and people/programs can pull out bits and pieces as

I've gone with a breakend-level GRanges notation as all (non-CNV) SV events
can be converted to this notation. CNV are not supported as they make a
claim about DNA segments, whereas the other events are making claims about
DNA segment adjacency - something that unfortunately the VCF specifications
does not distinguish between.


	[[alternative HTML version deleted]]

More information about the Bioc-devel mailing list