Large (Mb-scale) SVs

[yuliamostovoy]

I've been looking at the minigraph generated by HPRC, focusing especially on large structural variants in segmental duplications where we have strong evidence from orthogonal data that these variants are real. I've been impressed at how well the graph is capturing variants like insertions, deletions, and inversions in the ~10kb-100kb range. My question is about larger variants at the >1Mb scale. One specific example is at 7q11.23, where there is an inversion of ~2Mb between the two SD blocks. It has a population prevalence of ~5% and is present in some of the HPRC assemblies. Here is a screenshot of this region from hprc-v1.0-minigraph-grch38.gfa, with the SD duplicons colored - you can see how there are two large blocks consisting of several different duplicons:

The graph accurately represents variable CNVs in the two yellow segments, but no large inversions between the blocks. There's a similar issue at 16p12.2, which has an extremely common large inversion that isn't present in the graph. Do you know if there are any settings that could be changed to add these variants to the graph? Any chance that the newer version of minigraph would do a better job with them? I love the concept of minigraph and am very interested in applying it to these analyses. Thank you!