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

[eysung]

Hi everyone,
I'm trying to do feature-classifier classify-consensus-blast.
My command is:
qiime feature-classifier classify-consensus-blast --i-query INPUT_repseq.qza --i-reference-reads REFSEQ.qza --i-reference-taxonomy REFTAX.qza --o-classification OUT_classification.qza --o-search-results OUT_search-results.qza

More than half ran well, but few of them showed error below:
Plugin error from feature-classifier:

BLAST6 file is empty.

Debug info has been saved to /tmp/qiime2-q2cli-err-tp9ji65z.log

For more information,

• INPUT_repseq.qza is from after dada2 denoised-single
• Full process is like this:
  raw data - cutadapt - merge - (q2) import - denoising - taxonomy assignment

Thanks for your help,

EY

[Nicholas_Bokulich]

Hi @eysung ,

This error message indicates that the BLAST6 output is empty, because no hits were found. You should check the query, reference database, and parameters to make sure that these settings are appropriate for your intended action.

Good luck!

[eysung]

Hi, @Nicholas_Bokulich! Thanks for your reply

I tried this with other database to check the problem is reference database or not,
used data which this error happened, and brought the same error.

In this case, should I create data(repseq after denoising) again?
Also, if I want to check the query, do I have to export the repseq and see on the qiime viewer? Or is there any other shortcuts to check this?

EY

[Nicholas_Bokulich]

Hi @eysung ,

@eysung

In this case, should I create data(repseq after denoising) again?

No that is not necessary, unless if you detect some problem with how that file was generated. Instead, you may want to reduce the percent identity threshold or check other parameters.

@eysung

Also, if I want to check the query, do I have to export the repseq and see on the qiime viewer? Or is there any other shortcuts to check this?

try qiime feature-table tabulate-seqs

Good luck!

[eysung]

Hi @Nicholas_Bokulich,
I tried your advise, but it doesn't work well.

@Nicholas_Bokulich

Instead, you may want to reduce the percent identity threshold or check other parameters.

I have done this (before, I used the percent identity value as default) 0.7, but it still saying "BLAST6 file is empty".
I checked the query, and even the reference sequence.
I used three databases, EzBioCloud / Greengenes2 / Silva.

I'm trying 3 methods: consensus-blast, consensus-vsearch, sklearn,
I get most of the problems with consensus blast, very little with vsearch, and none with sklearn. I'm wondering why this is happening.

As an example, I exported Silva's reference sequence (V4). The only thing I noticed is that the length of the error file is 435 and the length of the reference is 253.
Is this causing the problem?

EY

[Nicholas_Bokulich]

Hi @eysung ,
Did you ever check the query sequences? This should always be the first thing. This error is telling you that there are no similar matches in the database, and you are now getting this error with 3 databases and 3 different methods. There is most likely something wrong with the query sequences.

Good luck!