Qiime demux summarize error with NextSeq data

[ebolyen (Evan Bolyen)]

Hi @ylor,

@ylor

Because I am looking at fish communities, I cannot use pre-trained classifers such as those from SILVA or GreenGenes. What would you suggest I use or do? Should I train my own feature classifiers?

Do you have some reference sequences (and their associated taxonomy)? Also, what is your target amplicon here? There are some other options that don't require training a classifier like:

  classify-consensus-blast    BLAST+ consensus taxonomy classifier
  classify-consensus-vsearch  VSEARCH consensus taxonomy classifier

@ylor

I would also like to run otu picking then use qiime feature-classifier classify-consensus-blast to blast my results before doing taxonomy assignment as well so that I can compare my results from dada2 denoising and otu picking using QIIME2. But again, I am not sure how to format my data properly to start otu picking.

Just to clarify, classify-consensus-blast is taxonomy assignment. As far as comparing OTU picking this tutorial is what you'll want to look at.

Since the input is just a feature-table and some representative sequences, you actually have two ways you can do this:

• You can either do dereplication (and chimera checking) directly on your SampleData[Sequences...] (similar to QIIME 1).

• or you can perform OTU picking on your DADA2 ASVs.

The two aren't mutually exclusive either, so you could try both as well.

Let me know if that helps!

[ylor (Yer Lor)]

Thanks @ebolyen.

I figured as much that those would be the possible routes for me to take. Having never generated my own reference sequences and taxonomy file, I am not sure what is all required in each of these files. Is there a specific format that the reference sequences need to be in? Can it just be a file containing fasta sequences downloaded from NCBI? How would I generate a corresponding taxonomy file? Can you provide me a little more guidance on how to generate properly generate these files so they are QIIME2 compatible?

Also, I am using several markers: 12s, 16s, coi and cytb. Would it be easier to generate a separate reference database per gene?

Thank you for bearing with me and answering all my questions!