I know there is not a lot of support for minigraph P-lines (#27) and it would break if a user added any new assemblies, but tools like vg deconstruct appear to perform better with path information than without. We observed ~6% more SVs (and were generally validated against assembly-based calls) when including path information than without. Arguable this may be an issue with vg rather than minigraph not providing P-lines, but that is the current state of tools.
This code is based heavily on the mgutils.js merge, except takes in a sample file and the paste *bed from stdin to create P-lines for each sample based on the paths taken during --call. I haven't really worked with js before, but I believe this is fairly streamlined.
This has a few obvious limitations.
minigraph call does not always yield the actual paths contributed by an assembly
"." alignments are basically treated as just skipping from start -> end node without any special consideration, and so are basically indistinguishable from "*" deletion alignments. (Maybe J-line would help?)
I know there is not a lot of support for minigraph P-lines (#27) and it would break if a user added any new assemblies, but tools like
vg deconstruct
appear to perform better with path information than without. We observed ~6% more SVs (and were generally validated against assembly-based calls) when including path information than without. Arguable this may be an issue withvg
rather than minigraph not providing P-lines, but that is the current state of tools.This code is based heavily on the
mgutils.js merge
, except takes in a sample file and thepaste *bed
from stdin to create P-lines for each sample based on the paths taken during--call
. I haven't really worked with js before, but I believe this is fairly streamlined.This has a few obvious limitations.
But overall we found this useful in our work (https://www.biorxiv.org/content/10.1101/2022.09.17.508368v1, page 4, some of the supplementary figures), so others may too while there is still a lot of graph <-> vcf exchange.
Best, Alex