Inconsistency between sourmash lca classify and sourmash lca gather

[ctSkennerton]

I created a signature and sourmash lca classify gives a particular genus, Eubacterium, but when I run sourmash lca gather with the same signature and database is not from a genome from Eubacterium. I was wondering if the "1 equal match" of the top match from sourmash lca gather was from a Eubacterium and if there was any way I could find out?

sourmash lca classify --query scaffolds_min1000.fasta.a.sig --db ~/Downloads/genbank-k31.lca.json.gz 
== This is sourmash version 2.0.1. ==
== Please cite Brown and Irber (2016), doi:10.21105/joss.00027. ==

loaded 1 LCA databases. ksize=31, scaled=10000
finding query signatures...
outputting classifications to stdout
ID,status,superkingdom,phylum,class,order,family,genus,species,strain
scaffolds_min1000.fasta.a,disagree,Bacteria,Firmicutes,Clostridia,Clostridiales,Eubacteriaceae,Eubacterium,,

 sourmash lca gather scaffolds_min1000.fasta.a.sig ~/Downloads/genbank-k31.lca.json.gz 
== This is sourmash version 2.0.1. ==
== Please cite Brown and Irber (2016), doi:10.21105/joss.00027. ==

loaded 1 LCA databases. ksize=31, scaled=10000
loaded query: /Users/connor.skennerton/Downl... (k=31)

overlap     p_query p_match 
---------   ------- --------
3.6 Mbp      77.8%   79.6%      Clostridium [Butyribacterium] methylotrophicum (** 1 equal matches)
320.0 kbp     6.9%    6.9%      Eubacterium callanderi
130.0 kbp     2.8%    2.8%      Eubacterium limosum

[ctb]

hi @ctSkennerton, good question! There is no easy (command line) way to figure this out... the slightly hacky way to do it today would be to use sourmash search to pull out all the matches, and then check against them with gather. We could add --save-matches to lca gather as well, and/or support output of equal matches to --csv.

[ctb]

#1310 and #1226 will make this much easier.

[ctb]

sourmash prefetch (#1370, released in sourmash 4.1.0) makes this straightforward!