Rely on Nextclade for ancestral amino acid inference

[j23414]

Context

slack conversation

I also noticed something kind of strange poking around with some mutations. For VP1 site 119, many internal branches are inferred to be M, yet I don’t see a single tip with M (blue branches here: https://next.nextstrain.org/staging/norovirus/all/genome?c=gt-VP1_119&m=div). I’m not sure how widespread this is

Good catch on the internal node state issue! I suspect that happens because the workflow translates the inferred ancestral nucleotide sequences for each internal node. An alternate approach that might fix this issue would be to pass amino acid sequences from Nextclade to the augur ancestral command (like we do here in seasonal flu) which would infer ancestral amino acid sequences per internal node. This approach should never be able to infer a state that doesn’t exist in the alphabet of the input sequences.

Description

Examples

• seasonal-flu

Possible solution

(Optional)