Cutadapt trim-paired odd result

[mayo]

Hello, I am new user of qiime2. I am trying to analyze my data. After multiplexed my data I trimmed it, but I got an strange result for the reverse sequences. Near to 230 pb, it appear reads with low quality and compact (demux-trimmed). I don understand why. I assumed that if I trim the sequences they would have had better quality. I really appreciate any advise. Should I work with the trimmed or demultiplexed sequences?

qiime cutadapt trim-paired
--i-demultiplexed-sequences demux.qza
--p-front-f CCTACGGGNGGCWGCAG
--p-front-r GACTACHVGGGTATCTAATCC
--verbose
--o-trimmed-sequences trimmed.qza

demux.qzv (3.3 MB)

[llenzi (Luca)]

Hi Mayo,
I see something similar, and I'm sharing with you my thought on this, hopeful any expert may support the conclusion.
Using the command you sent, same I use, in fact we trim from the left side of the sequences (left part, front), hence the low quality tail will remain there.
My thinking is that, given the primer may be at any position in the initial part of the sequence, the overall effect is to dis-align the sequences bases. That is, two bases that are at the same position in the raw reads may be not be at the same position after cutadapt trimming.
In my mind, it means that sequence cycle 220, with its potentially low quality, may be at position 200 for a trimmed sequence and at position 220 for a sequence in which cutadapt can not find a primer match. The visualiser still visualise the quality from position 1, of each sequences and that creates a heterogeneity of quality scores (from different sequencing cycles) at the tail of the plot.
Hope it makes sense. Please, any expert correct me if I'm wrong!
Luca

[thermokarst]

Hello @mayo!

@mayo

Near to 230 pb, it appear reads with low quality and compact (demux-trimmed)

Have you compared the quality plots before and after trimming? That seems like the first place to start here.

@mayo

I don understand why. I assumed that if I trim the sequences they would have had better quality

This result doesn't strike me as odd --- you didn't do any trimming based on quality score, just removal of non-biological sequence...

@mayo

Should I work with the trimmed or demultiplexed sequences?

If you are going to use a modern denoiser like DADA2 or deblur you will need to remove all non-biological sequences from your reads.

:qiime2:

[mayo]

Dear thermokarst, thank you for your reply. Yes, I compared both files before and after trimming.

@thermokarst

Have you compared the quality plots before and after trimming?

Before trimmed

After trimmed

Thank you for your advise.

[thermokarst]

@mayo, can you share the two QZV files, instead of screenshots?

[mayo]

Hello thermokarst. Sure.
demux.qzv (288.9 KB)

demux-trim.qzv (292.7 KB)

Thank you again for your advice.

[thermokarst]

@mayo - perfect, thanks!

Okay, this makes sense to me after interactively looking at the plot! The reason you see such a stark decrease in quality on the reverse reads is because not all of your reads are being trimmed. Most of your reverse reads are being trimmed to 230-231 nts long, but it looks like ~100 reads are still 251 nts long. Maybe you could try bumping your trim error rate up to something higher than the 10% default value? Alternatively, just trunc the rev reads at ~230 nts if running q2-dada2.

Keep us posted!

PS - the same thing is happening on your forward reads, about 100 reads are still the original length.