Genome build

[jameshadfield]

The included strains are taken by finding a part of the tree I'm confident doesn't
have reassortment in from looking at the segment trees.

The genome annotation file was generated by concatenating the S,M & L segments
from the Nextclade genbank references in https://github.com/nextstrain/andv

While the clock signal didn't look terrible, the timetree inference placed the CA of the Hondius outbreak back in 2004 so I've removed the timetree.

I used a new snakefile, but we can add these to the main snakefile if preferred. Run via:

snakemake --cores 1 --snakefile genome.snakefile -fp

TODO

• There's a lot of spurious mutations at segment boundaries which aren't being correctly masked. I'll add a script to mask these blocks for tree/refine steps. This should also improve the temporal signal.

[jameshadfield]

My main concern is the rooting of the Hondius outbreak subtree. The (putative?) index case (PP_006WDKH, South Africa, 2026-04-26) has 2 unique mutations relative to the others and in our rooting this places it as a descendant. One of them is ~52nt from a segment boundary, so maybe that should be masked, leaving only a G8979 mutation. There's too few strains for TImeTree temporal analysis, but setting PP_006WDKH as the root lines up better with the metadata.