--match-per-kmer increasing indefinitely

[sean-workman]

Hello Jaebom,

I am running into a strange problem using the Refseq Prokaryote/Viral DB to try and classify paired-end short reads. As the run goes on, it just keeps increasing the --match-per-kmer parameter and is unable to continue:

classify fastp_def/CMAB71413_OST_fastp_R1.fastq.gz fastp_def/CMAB71413_OST_fastp_R2.fastq.gz /home/sdwork/scratch/metabuli_dbs/refseq_prokaryote_virus me
tabuli_prokaryote_viral CMAB71413_OST --threads 32 --max-ram 150

Metabuli Version (commit):                            	77f39b659ec041699f72f15636655ee9500f7642
Threads                                               	32
Sequencing type                                       	2
Min. sequence similarity score                        	0
Min. query coverage                                   	0
Min. num. of cons. matches for non-euk. classification	4
Min. num. of cons. matches for euk. classification    	9
Min. score for species- or lower-level classification.	0
Allowed extra Hamming distance                        	0
Directory where the taxonomy dump files are stored
Mask residues                                         	0
Mask residues probability                             	0.9
RAM usage in GiB                                      	150
Number of matches per query k-mer.                    	4
Accession-level DB build/search                       	0
Best * --tie-ratio is considered as a tie             	0.95

DB name: 1287220998
DB creation date: 2024-4-1
Loading the list for taxonomy IDs ... Done
Indexing query file ...Done
Total number of sequences: 57358948
Total read length: 12407825166nt
Extracting query metamers ...
Time spent for metamer extraction: 19
Sorting query metamer list ...
Time spent for sorting query metamer list: 16
Comparing query and reference metamers...
--match-per-kmer was increased to 8 and searching again...
Extracting query metamers ...
Time spent for metamer extraction: 9
Sorting query metamer list ...
Time spent for sorting query metamer list: 9
Comparing query and reference metamers...
--match-per-kmer was increased to 12 and searching again...
Extracting query metamers ...
Time spent for metamer extraction: 6
Sorting query metamer list ...
Time spent for sorting query metamer list: 6
Comparing query and reference metamers...
--match-per-kmer was increased to 16 and searching again...
Extracting query metamers ...
Time spent for metamer extraction: 5
Sorting query metamer list ...
Time spent for sorting query metamer list: 4
Comparing query and reference metamers...
--match-per-kmer was increased to 20 and searching again...

.....

--match-per-kmer was increased to 144 and searching again...
Extracting query metamers ...
Time spent for metamer extraction: 1
Sorting query metamer list ...
Time spent for sorting query metamer list: 0
Comparing query and reference metamers...

These reads are able to be classified by the Refseq Viral DB, as well as the ICTV DB using your personal fork of Metabuli (#96). The time I had requested from our cluster ran out while the parameter was at 144. Any thoughts on what could be causing this?

Thanks!

[jaebeom-kim]

Hi Sean!
Thank you for reaching out!
My personal fork mentioned here is currently compatible only with the ICTV DB.
I think the result from using refseq-viral DB must be errorous.
Please try the latest release in the main repository :)

[sean-workman]

Oh yikes haha I assumed that working with the ICTV database was an additional feature. Will go give that a go, thanks!

[sean-workman]

Oh - it would seem I am using the appropriate Metabuli.

metabuli Version: dc985814afbcbd2fda50878138165ea3006ce4f1

My Slurm submission script is identical to the one I successfully used with the Refseq Viral DB, except of course for changing which DB I am using.

[jaebeom-kim]

Hmm.. The version written in the first comment of this issue is 77f39b659ec041699f72f15636655ee9500f7642.

Metabuli Version (commit): 77f39b659ec041699f72f15636655ee9500f7642.

If you still encounter an error with dc985814afbcbd2fda50878138165ea3006ce4f1, could you provide the sample to reproduce the error?