Removal of the superkingdom rank in NCBI taxonomy breaks MEDI

[nayeon1994]

Hello,

I'm currently using medi according the instructions on Github.
However, I've encountered some errors.

The error occurs during the process architeuthis_filter step in quant.nf.
If I exclude this step, the workflow runs without errors.

To debug the issue, I started by looking into the database.
I noticed that when I run the following command:

architeuthis mapping filter {kraken.k2} --data-dir {medi_db/taxonomy} # (with options like --max-entropy 1 --max-multiplicity 10 --min-consistency 0.1)

no reads are retained. I tried adjusting the parameters with various thresholds, but still, no reads are retained.

When I run :

architeuthis mapping score {kraken.k2} --data-dir {medi_db/taxonomy} -f "{p};{c};{o};{f};{g};{s}"

it works correctly.

But if I run:

architeuthis mapping score {kraken.k2} --data-dir {medi_db/taxonomy} -f "{k}"

or

architeuthis mapping score {kraken.k2} --data-dir {medi_db/taxonomy} -f "{k};{p};{c};{o};{f};{g};{s}" #default

no reads are ratined again.

This leads me to believe there might be an issue with the database, specifically with the kingdom-level information.
I checked the manifest.csv file in the medi_db directory and found that the kingdom names for all species are empty (NaN).

Could you help me debug this issue?

Thank you!

[cdiener]

Hmm weird. I also observed the issue with the NA kingdom rank which is due to a change in the NCBI taxonomy. It already fixed it in the development version. However, that should not affect the scoring but maybe there is some odd side effect with the missing rank. I will investigate a bit further.

For now you could try to substitute {k} with {K}:

architeuthis mapping score {kraken.k2} --data-dir {medi_db/taxonomy} -f "{K};{p};{c};{o};{f};{g};{s}"

[nayeon1994]

Thanks for help me.

Your recommendation worked well - substituting {k} with {K} solved the issue.

Currently, I plan to use the architeuthis mapping score instead of the mapping filter, and then add a custom script to parse the score matrix in order to utilize the medi pipeline.

Right now, the mapping filter defaults to the format "{k};{p};{c};{o};{f};{g};{s}" without allowing any modifications by architeuthis users.
Therefore, adding a --format option to the mapping filter would be a significant improvement.

Thanks again!

[cdiener]

Hi, sorry for that. I will add an argument in architeuthis for this and change the default.

[cdiener]

I fixed the issue in architeuthis. It also defaults to the kingdom now, like in the comment and it also allows setting the format in filtering. Should be on bioconda soon, it's already in the queue.

[atyakhtmpg]

Hi Christian. I ran into the same bug - thanks for the quick fix - but the downstream processing still produces an error:
` >> Checking for Valid Options...

Running Bracken
/SKIPPED/bin/bracken: line 98: [: S/sample_ID_bracken.tsv: unary operator expected
>> python src/est_abundance.py -i S/sample_ID.tsv -o S/sample_ID.b2 -k /SKIPPED/medi_db/database150mers.kmer_distrib -l S -t 10
PROGRAM START TIME: 04-16-2025 14:14:35
Traceback (most recent call last):
File "/SKIPPED/medi2025/bin/src/est_abundance.py", line 488, in
main()
File "/SKIPPED/medi2025/bin/src/est_abundance.py", line 271, in main
elif main_lvls.index(level_id[0]) >= branch_lvl:
ValueError: '2' is not in list`

My guess is that the problem is caused by the lines in the input .tsv that had "2" instead of "R2" in the 4th position - specifically for my example sample, four of them - Bacteria, Archaea, Eukaryota and Duplodnaviria (the set might vary from sample to sample). Once I manually replaced those, the script completed without an error. Please could you check that?

[cdiener]

Good point. The removal of the superkingdom rank affects more steps in the pipeline and breaks a bunch of tools since it was a canonical rank. I will see if I can put a workaround in place in fixk2repor.R while I fix the root issue. If one of you can share their report file that would help to debug.

[cdiener]

Okay should be fixed in version 0.2.1 for now. The consistence filtering starts from the kingdom for now. After a new taxonkit including the fix from shenwei356/taxonkit#117 is released, it will be returned to the domain.

[atyakhtmpg]

Tested; works in my case. Thanks!

[cdiener]

Here is the official announcement. Probably needs another workaround for Viruses.