display ANI in search results?

sourmash-bio / sourmash

Quickly search, compare, and analyze genomic and metagenomic data sets.

http://sourmash.readthedocs.io/en/latest/

Other

466 stars 79 forks source link

Open ctb opened 2 years ago

ctb commented 2 years ago

Once https://github.com/sourmash-bio/sourmash/pull/1967 is merged, ANI will available in CSV files! 🎉

It is also available in sourmash compare matrix output, if --ani is used. 🎉

I don't think it is displayed anywhere else.

Do we want to add ANI to the search output, @bluegenes? I'm in favor - the search results are pretty sparse so I think we even have room for them.

Not sure about gather, though. I think the k-mer overlap approaches makes more sense, maybe? But it would be nice to have as an option, maybe? 🤔

Maybe do it for search first, since I'm pretty sure that's a good idea, and then a separate PR (no hurry) for gather?

bluegenes commented 2 years ago

ANI to search output would be great, but a few thoughts:

we can't estimate ANI for num signatures, so num vs scaled outputs would be different (maybe they already are?)
search uses Jaccard by default, which isn't as good as containment for ANI. Does adding ANI to the search output encourage use of jaccard--> ANI?
How do we want to handle excessive jaccard error? Currently emit a warning, do we want to zero out?
Do we want to allow ANI estimation for abund searches (ANI ignores abundance)? If so, from jaccard?

ctb commented 2 years ago

we can't estimate ANI for num signatures, so num vs scaled outputs would be different (maybe they already are?)

ahh, I'd kinda forgotten that.

bluegenes commented 2 years ago

ahh, I'd kinda forgotten that.

we could actually estimate ANI from num sketches using the Mash Distance, but I'd rather not, because:

ctb commented 2 years ago

agreed.