Results deblur stats

Please tell me why in some of my samples a large number of readings are in groups unique-reads-missed-reference and reads-missed-reference? How to fix it? With taxonomic classification, the remaining sequences are identified as 2 organisms, although there should be a large variety.
Thanks for your reply!

Hello Yulia,

Welcome to the forums! :qiime2: Thank you for your patience.

The Deblur pipeline relies on a reference, and it looks like most of the reads in samples S4-16 and S1-16 did not match the reference. There's a bunch of reasons this could happen...

What region did you amplify? What database are you using for reference? What settings did you use to preprocess you data and run deblur?

I amplified V3-V4 region and used Greengenes Database, also used qiime vsearch join-pairs, qiime quality-filter q-score and qiime deblur denoise-16S.

qiime vsearch join-pairs
--i-demultiplexed-seqs /the path/
--o-joined-sequences /the path/

qiime quality-filter q-score
--i-demux /the path/
--o-filtered-sequences /the path/
--o-filter-stats /the path/

qiime deblur denoise-16S
--i-demultiplexed-seqs /the path/
--p-trim-length 0
--p-jobs-to-start 4
--o-stats /the path/
--o-representative-sequences /the path/
--o-table /the path/

Hello Yulia,

That pipeline looks good to me! My only thought is that the default vsearch join-pairs settings might not work well for your read length and V3-V4. (Did you use 300 or 600 cycle kits?)

I feel like this under-classification is an important clue, but I'm not sure what to make of it.

Let's see if @thermokarst has any ideas about deblur's performance.


I used 600 cycle kits. Previously, for other samples obtained in a similar way, I also used this pipeline and there were no problems.

I had the idea that the sample was contaminated during the amplification step and therefore the wrong sequences were obtained. Is it possible?

1 Like

It's totally possible! I was approaching this with the assumption that the data was fine and something went wrong with the pipeline. However, these could be low quality / contaminated samples.

Would you be will to post the artifact so we could see the provenance?

Yes, sure!
I hope this helps to avoid this problem in the future.

demux-joined-filter-stats.qza (16.3 KB)

Hi @Yulia_Kocharovskaya,

I'm going to step in for @colinbrislawn, if it's okay with him, and try to do some trouble shooting as well.

I have two or three potential thoughts.

  1. Did you trim your primers? Did someone else remove your primers? You might try using cutadapt to trim and discard as a quality check.
  2. I'm assuming you quality filtered ahead of time?
  3. What taxa are they hitting? Is there a possibility of mislabeling anywhere along the pipeline (its more common than you might think)


1 Like