We are getting closer to being able to separate out the lineage stuff from the revindex stuff in LCA, so we could envision combining searches on signatures and SBTs with taxonomic reporting (where available).

The basic idea here is that we combine taxonomy with search and gather on signatures/SBTs/LCAs, by connecting identifiers with lineages -- do dynamically what sourmash lca index does once.

This falls under use cases of taxonomic filtering and reporting.

What I don't know how to do is actually apply it in reality. A clunky (but perhaps functional) way that I'd been idly thinking of would be to build a taxonomy command-line interface that ingested the CSVs output by sourmash search and gather and did various kinds of taxonomic reporting and manipulation on them, when given a taxonomy or lineage database.

I don't really want to rewrite / extend gather to also have taxonomy... ugly complex code.

It might be possible to add taxonomy/lineage-aware selector functionality on signatures and databases, e.g. "run a gather on this database, but ignore all results that are not Archaea."

Should also look at sourmash lca gather to see about collapsing results taxonomically...

Thoughts welcome!

related conversations

• the ZipStorage work in [MRG] add ZipStorage, support loading tree from storage #648 makes it plausible to package SBTs up together with a lineage.
• related conversation in this comment in #581