vg call coverage/suppport

Question

[Danny]

Hello,

I am using vg 1.48.0 to try and calculate SV frequencies from short-read poolseq data mapped to a pangenome using giraffe. I was wondering about how the read depth for the reference vs alternate allele is calculated in vg call - is it the (total bp mapped)/(length of path) or (total bp mapped)/(length of path mapped). I have filtered for mapping quality, so I am wondering if one path had a repetitive region present elsewhere in the genome, this path would get less coverage (and for single sample data, be less likely to be called)? I wasn't sure from Hickey et al. 2020 but apologies if I've missed this somewhere.

Best wishes,

Daniel Wood

Answer 1

[glenn.hickey]

For SV's it's looking at the average (base by base) read coverage across the entirety of each allele (reference or alt treated the same) in the site. If two alleles share a node, its coverage is split evenly between the two alleles. For smaller variants (<50bp), it uses the minimum instead of average coverage.

[Danny]

Ok great, thanks for clarifying! So do you think filtering out low MAPQ reads might cause an issue with SV calling, if one allele has more repetitive sequence than the other?