I used this command- qiime dada2 denoise-paired --p-trunc-len-f 225 --p-trunc-len-r 225 --i-demultiplexed-seqs paired-end-demux.qza --o-representative-sequences reporepseq.qza --o-table repotable.qza --o-denoising-stats dadastats.qza
Here is the demux summary file which shows generally good quality- demux.qza.qzv (318.2 KB)
And here is the dada2 summary qzv- dadastats.qzv (1.2 MB)
Looking at the samples on the dada summary qzv in numerical order, this very odd pattern forms--- 5 samples look good, 2 are intermediate, 5 are awful, and repeat over and over throughout the dataset.
Is this some weird artifact stemming from the sequencer? Something I did? I'd understand if some samples were junk, but this cyclical pattern is odd and difficult for me to understand.
I just went through your DADA stats visualisation file and it was certainly strange to see that 5/2/5 repeating pattern throughout your samples.
I don't think it was anything you did, but there may have been some issues during sequencing which created this strange pattern. Since it occurs in this 'batch pattern' it might have been something to do with how they were loaded onto the sequencer.
The truncation parameters you have used in your DADA2 command (--p-trunc-len-f(r) 225) might be having some effect too.
You could try re-running this step and adjusting these numbers, perhaps try 179 for the forward (quality seems to dip at position 183) and 220 for your reverse (quality dips at position 226).
It would be interesting to see if that pattern effect changes after this.
That is an extremely strange run for many small reasons.
Illumina's new binned quality scores: discussed here
Very even reads-per-sample
Min of 13k and max of 22k is fantastic! How did you get that to happen?
Blocks of sample losing 99% of reads during the quality filter
That last one is the big mystery!
A sequencing core at a previous employer had a strange issue like this.
We discovered a problem with the liquid-handling robot we used to do PCR.
I recommend lateral thinking.
Can you check the PCR plate layout?
Were these stored in a different freezer, run in a different thermocycler?
What primers did you use?
Thank you for your help on this matter and your multiple recommendations.
I can say that we used a commercial sequencer for this run so I am not completely privy to the specific layout of the plate or the freezers / thermocycler. I have reached out for that information but have not gotten a response yet. Likewise, I have asked about the even values amongst so many of the samples to see if that is in their norm.
We did use V4V5 primers, 515F 907R.
Per Mike Stevenson's comment, I did truncate at a couple different lengths and got much better results on the dada2 output. If I learn anything from the company or observe anything odd in the workflow on my end as I move forward, I will update.
