Deblur empty output for two samples

[serugra (Sergio Rubio)]

Hi, I just performed a deblur analysis using the 2019.7 version and after:

qiime deblur denoise-16S \
--i-demultiplexed-seqs demux-joined-filtered.qza \
--p-jobs-to-start 8 \
--p-trim-length 400 \
--p-sample-stats \
--output-dir deblur

qiime deblur visualize-stats \
--i-deblur-stats deblur/stats.qza \
--o-visualization deblur/stats.qzv

When checking the stats.qzv artifact two of my samples have all the reads filtered and no output in deblur.

stats.qzv (212.7 KB)

The reads-raw value for them is much higher than the rest of the samples. Could someone explain me or at least hypothesize what went wrong?

Also, we recently changed the polymerase we are using and have noticed the reads-hit-reference/reads-raw is around 10% for most samples. Is this a normal value? I have checked previous runs with another polymerase and this value is closer to 30%.

Thanks

[wasade (Daniel McDonald)]

Hi @serugra,

Sorry for the very slow reply here, this slipped accidentally.

We typically see much higher percentages hit the positive filter. First though, you’re seeing higher than 10%: the fraction-missed-reference is close to zero for most samples (so what makes it through seem to be of the target amplicon type). There is a very large number of artifactual reads (adapter, phix, etc) and/or singletons though – the first sample is around 70%.

I recommend reducing the trim length as that will contribute to a large number of singletons. My guess is the 70% above is probably driven by singletons. If these are 515F/806R, then often just the first 90nt are sufficient for many studies and questions.

It is weird that two samples are empty, with a nan. I think that can only occur if nothing makes it through Deblur. There is a lot more error in longer reads, and I suspect that is also a factor here.

All the best,
Daniel