Update ingest

[genehack]

No description provided.

[corneliusroemer]

@genehack did you name this PR injest in jest? 😄 I renamed to the more canonical ingest, I hope you forgive me :)

[genehack]

@genehack did you name this PR injest in jest? 😄 I renamed to the more canonical ingest, I hope you forgive me :)

thanks, no, not intentionally, my brain gest 😁 really wants that soft g to be a j ...

[genehack]

Bringing comments from @jameshadfield over from Slack for ease of change-tracking:

Notes after briefly playing with the tree:

• If you can track down an annotation (of the CDSs after cleavage from the polyprotein) that would really enhance the diversity panel
• Looks like there may be seven genotypes in the literature: "West African genotypes I and II, East African genotype, East/central African genotype, Angolan genotype, and South American genotypes I and II" . Assuming these fall nicely on the tree (and they should - they're actual genotypes not serotypes) then labelling these via augur clades would be helpful.
• The clock signal looks weak (view the tree in "clock" layout). Some of the clades look like they may have a stronger signal, so potentially we could run it for individual genotypes. There may be interesting biology here as the rates look different in different clades, but I haven't read the literature, this is only from interacting with the tree.
• Assuming we remove the timetree, we can still include isolation date as metadata, but it doesn't work well as a colouring (Auspice shortcoming) - in the past I've encoded a continuous colouring of year which Auspice does display nicely.
• I think it's valid to run augur traits on region. Potentially country too, but that looks like there's more sampling bias (or maybe not, I don't know the prevalence).
• [minor] lat/longs missing for Guinea-Bissau, Côte d'Ivoire

[genehack]

• If you can track down an annotation (of the CDSs after cleavage from the polyprotein) that would really enhance the diversity panel

Yeah, I found the RefSeq for yellow fever virus and it's got a complete breakdown, so I can do this.

• Looks like there may be seven genotypes in the literature: "West African genotypes I and II, East African genotype, East/central African genotype, Angolan genotype, and South American genotypes I and II" . Assuming these fall nicely on the tree (and they should - they're actual genotypes not serotypes) then labelling these via augur clades would be helpful.

👍

• The clock signal looks weak (view the tree in "clock" layout). Some of the clades look like they may have a stronger signal, so potentially we could run it for individual genotypes. There may be interesting biology here as the rates look different in different clades, but I haven't read the literature, this is only from interacting with the tree.

I'm hearing "maybe down the road but not critical path" here, push back if that's wrong.

• Assuming we remove the timetree, we can still include isolation date as metadata, but it doesn't work well as a colouring (Auspice shortcoming) - in the past I've encoded a continuous colouring of year which Auspice does display nicely.

kk.

• I think it's valid to run augur traits on region. Potentially country too, but that looks like there's more sampling bias (or maybe not, I don't know the prevalence).

👍

• [minor] lat/longs missing for Guinea-Bissau, Côte d'Ivoire

yep, saw that error during build, thanks for highlight.

[jameshadfield]

Some of the clades look like they may have a stronger signal, so potentially we could run it for individual genotypes.

I'm hearing "maybe down the road but not critical path" here, push back if that's wrong.

💯 - not something for the initial build!

Looking at this again today there are signs of recombination in the tree, although the literature suggest[s] that recombination between YFV strains is unlikely. The high number of homoplasies on the deep branches, and the clustering of them, sure looks consistent with recombination - e.g. 5389G,5437T,5443T,5485A,5497T. If this is more prevalent deeper in the tree it would result in less clock signal, or said another way the clock signal might still be ok for individual YF-genotypes.

It'd be useful if Auspice were able to show a colouring for unique mutation counts, homoplasic counts and the ratio. I'll look into this at some point. Could even use these as branch metrics if we get around to nextstrain/auspice#1769.

[j23414]

Tested ingest again, works for me!

[joverlee521]

Just noticed this repo doesn't auto-delete merged branches

Clicked the little box for this repo