Unable to set only one haplotype of a diploid to reference-sense in gbwt

[ASLeonard]

Hi,

Setting the "references" via the GFA RS:Z: tag in the header or using vg gbwt --set-reference for a given panSN sample name works fine. However, for triobinned assemblies (particularly in agriculture), one haplotype might be taken as the reference while the other is not. Setting the header with RS:Z:sample#hap results in HAPLOTYPE-sense, while trying vg gbwt --set-reference "sample#hap" fails due to a prohibited character ("#").

Perhaps the easiest solution is to promote the entire sample to reference and just ignore that the non-reference haplotype is REFERENCE-sense, but do you have any other ideas? I guess handling a half-reference half-haplotype sample might complicate a lot of the internal workings for diploid sampling, but parsing the haplotype field of panSN for setting reference paths seems doable.

This is using v1.63.1.

Best,
Alex

[jltsiren]

Only entire samples can be set as references.

GBZ could in principle specify references at the level of individual paths, but there are no interfaces for passing that information to the graph. Not on the command line, not in the API, and not in the file formats. Some other implementations determine the sense by parsing the path name, and renaming the paths is the only way to switch between haplotype and reference senses. I'm not exactly sure about the rules. sample#sequence is clearly a reference and sample#haplotype#sequence#something is clearly a haplotype, but I'm not sure how PanSN names will be interpreted.