what does 'disagree' mean in output?

[sapuizait]

Apologies if this is sth trivial or has been asked before but I could not find an answer in this forum or by googling it.

I use sourmash to classify genomes/MAGs and while I understand in the output table the successful classification is marked as 'found' and the unsuccessful as 'nomatch' - what does the 'disagree' mean? I assume that it means that matches were found but based on the software's cutoff it is not very happy? The matches are not great? Is that it? if yes, can I still use the result? How reliable is it?

thanks

here is an example:

2023_1030076_1_MG_127_23112020_S0_L001bin.47.fa,disagree,d__Bacteria,p__Bacteroidota,c__Bacteroidia,o__Bacteroidales,f__Barnesiellaceae,g__Barnesiella_A,,

2023_1030076_1_MG_127_23112020_S0_L001bin.4.fa,nomatch,,,,,,,,

2023_1030076_1_MG_127_23112020_S0_L001bin.52.fa,found,d__Bacteria,p__Spirochaetota,c__Spirochaetia,o__Treponematales,f__Treponemataceae,g__Treponema_D,s__Treponema_D sp900767955,

2023_1030076_1_MG_127_23112020_S0_L001bin.57.fa,found,d__Bacteria,p__Verrucomicrobiota,c__Verrucomicrobiae,o__Verrucomicrobiales,f__Akkermansiaceae,g__Akkermansia,,

[ctb]

great question!

I vaguely remembered writing something about it somewhere, but had to go digging 😆 . And it was by no means easy to find!

Here's what I wrote in this blog post:

Interpreting the CSV file

The CSV file has five columns: name, taxid, status, rank_info, and lineage. name is the name in the signature file; taxid and lineage are the NCBI taxonomic ID assigned to that name, and the taxonomic lineage of that ID.

The status and rank_info require a bit more explanation.

There are three possible status values at present: nomatch, found, and disagree.

nomatch is hopefully obvious - no (or insufficient) taxonomic information was found for k-mers in that signature, so you get a taxid of 0, and an empty rank and lineage:

TARA_IOS_MAG_00005,0,nomatch,,

found is hopefully also pretty obvious: it says that a simple straightforward lineage was found across all of the databases, and gives you that lineage down to the most detailed rank found:

TARA_PSW_MAG_00129,28108,found,species,Bacteria;Proteobacteria;Gammaproteobacteria;Alteromonadales;Alteromonadaceae;Alteromonas;Alteromonas macleodii

disagree is the least clear. Here, the assigned rank is the rank immediately above where there is a taxonomic disagreement, and the taxid & lineage refer to the name at that rank (the least-common-ancestor at which an assignment can be made).

For example, look at this line in the CSV file:

TARA_ASW_MAG_00029,1224,disagree,phylum,Bacteria;Proteobacteria

TARA_ASW_MAG_00029 has k-mers that are shared between different orders: 'Pseudomonadales' and 'Rhodobacterales'. Therefore, the classifier status is disagree, and the classified taxid is at rank phylum - just above order.

---

make sense?

thanks again for asking!

Since you're using the LCA module, I also wanted to point you at this issue: #2760

Nowadays we suggest using sourmash gather to produce a CSV file containing a list of overlapping genomes, followed by sourmash tax to classify the genome or metagenome contents - it's proven to be much more accurate (specific) than our LCA approaches. See #2699 for a tutorial demonstrating it on both metagenomes and genomes!

If you have any more questions, ask away! And please do leave this issue open so I can update the docs appropriately.

[sapuizait]

WOW, thanks a lot for the detailed answer! And thanks for pointing me to the gather > tax approach. For the record, I am very happy with the classification because the 'nomatch', if I try to blast them for example, they only have some very distant hits with some uncutured/unclassified isolates that nobody knows what they are :D
Again, thanks for the great software!