Plugin error from feature-classifier: BLAST6 file is empty.

[cefas23]

Hi, I'm trying to build updated databases for 18S, COI and 12S (mifish).

Once I have my fasta file with the sequences and the taxonomy tsv files i use qiime commands to convert into qza files:

qiime tools import --type 'FeatureData[Sequence]' --input-path xxx.fasta --input-format DNAFASTAFormat --output-path xxx.qza
qiime tools import --type 'FeatureData[Taxonomy]' --input-format HeaderlessTSVTaxonomyFormat --input-path** xxx.tsv --output-path xxx.qza**

I then use the feature classifier to assign taxonomies:
qiime feature-classifier classify-consensus-blast
--i-query mifish_rep-seqs.qza
--i-reference-reads references.12s.miyaSeqs.qza
--i-reference-taxonomy references.12s.miyaTaxonomy.qza
--p-perc-identity 0.97
--output-dir references.12s.miyaTaxonomy97

But I always get an error for any new database I try:
Plugin error from feature-classifier:

BLAST6 file is empty.

Could you please help me out? This used to work really well, maybe 3-4 years ago but not anymore.
Thank you.
Vera

[colinvwood (Colin Wood)]

Hello @cefas23,

Could rerun with the --verbose flag and post the output? It seems like maybe it's erroring because there were no hits, thus an empty blast6 file.

[cefas23]

Hello

I've checked and that might be the reason. thanks.
another error with a different data and database shows up.
When I run :
qiime feature-classifier classify-consensus-blast
--i-query nis_trimmed_ASVs/nis_rep-seqs.qza
--i-reference-reads COIsequence23.qza
--i-reference-taxonomy COItaxonomy23.qza
--p-perc-identity 0.97
--output-dir nis_COItaxonomy97/
--verbose

I get this error:
(1/1) Invalid value for '--i-reference-reads': COIsequence23.qza was created
by 'QIIME 2023.5.1'. The currently installed framework cannot interpret
archive version '6'.

could you please advice?
Thank you.

[colinvwood (Colin Wood)]

Hello @cefas23,

It looks like COIsequence23.qza was generated with qiime 2023.5, which introduced a new archive version, but you're using something lower than 2023.5. If you upgrade to 2023.5 or generate COIsequence23.qza with <2023.5, this issue should go away.

[cefas23]

Many thanks @colinvwood !
I managed to BLAst using that COIsequence23 database but the sequence names did not match the taxonomy. Had to re-do a new one. Upon using the classifier command I got a complete different error about using "Use pandas.concat instead." which I have no idea how to fix. Please see the command and message bellow please let me know if you have any idea how I should proceed? Thank you.

% qiime feature-classifier classify-consensus-blast \

--i-query nis_trimmed_ASVs/nis_rep-seqs.qza \

--i-reference-reads GenBank_seq_080319.qza \

--i-reference-taxonomy GenBank_tax_080319.qza \

--p-perc-identity 0.97 \

--output-dir nis_COItaxonomy97/ \

--verbose

Running external command line application. This may print messages to stdout and/or stderr.

The command being run is below. This command cannot be manually re-run as it will depend on temporary files that no longer exist.

Command: blastn -query /var/folders/jh/3nkmqc753rldc1pq_n6ytsfw0000gn/T/qiime2/cefas23/data/259f4921-8617-45c0-ae3a-b4b37e292f7b/data/dna-sequences.fasta -evalue 0.001 -strand both -outfmt 6 -subject /var/folders/jh/3nkmqc753rldc1pq_n6ytsfw0000gn/T/qiime2/cefas23/data/54e09e43-2e62-4a85-b889-09021d6258f2/data/dna-sequences.fasta -perc_identity 97.0 -qcov_hsp_perc 80.0 -max_target_seqs 10 -out /var/folders/jh/3nkmqc753rldc1pq_n6ytsfw0000gn/T/q2-BLAST6Format-q6da4uz7

Warning: [blastn] Subject_2926544 Y14709.1 Choanomphalus baicalensis mitochondrial cytochrome c oxidase subunit I gene, partial: Subject sequence contians no data

/Users/cefas23/miniforge3/envs/qiime2-2023.7/lib/python3.8/site-packages/q2_feature_classifier/_blast.py:80: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.

missed = missed.append(

/Users/cefas23/miniforge3/envs/qiime2-2023.7/lib/python3.8/site-packages/q2_feature_classifier/_blast.py:84: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.

result = result.append(missed, ignore_index=True)

[cefas23]

Hi @colinvwood
I thought it could be the new qiime2-2023.7 so I've installed qiime2-2023.5 and although I did not get the panda error, I got another error:

zsh: killed qiime feature-classifier classify-consensus-blast --i-query 0.97

I have no idea why this happens. could help please?
Thank you.

[colinvwood (Colin Wood)]

Hello @cefas23,

The pandas.concat messages are actually warnings, not errors.

As far as the zsh: killed error, could you run the command that triggered that with the --verbose flag and post the whole error output?

[cefas23]

thank you @colinvwood. the command was
~ % qiime feature-classifier classify-consensus-blast \

--i-query nis_trimmed_ASVs/nis_rep-seqs.qza \

--i-reference-reads COI-midori-sequences.qza \

--i-reference-taxonomy COI-midori-taxonomy.qza \

--p-perc-identity 0.97 \

--output-dir nis_midori_COItaxonomy97/ \

--verbose

but I cannot see the error output

[colinvwood (Colin Wood)]

Hello @cefas23,

What do you mean you can't see the output? Nothing is printed at all? Or only zsh: killed qiime feature-classifier classify-consensus-blast --i-query 0.97?

[cefas23]

yes indeed I can only see zsh: killed qiime feature-classifier classify-consensus-blast --i-query 0.97 nothing else!

[colinvwood (Colin Wood)]

Hello @cefas23,

It might be a memory issue, how long does the command run before it gets killed?

[cefas23]

Hello @colinvwood,
it runs probably around 1-2h but it's the first time it happens, perhaps because the new databases are bigger? could there be a way around this?
Thank you.

[colinvwood (Colin Wood)]

Hello @cefas23,

To confirm that it is a memory issue you could look at the memory usage while the process is running using a tool like htop or top.

As far as a way around this, if it does turn out to be a memory issue, then getting more memory will be the easiest solution. If you're on a cluster you can just request more, for example.

From my knowledge, this command shouldn't be very memory intensive though, so it could be a different issue.