ASVs vs 97% OTUs when using -usearch_global to create an OTU table

[Marita White]

Hello,

When using the following commands in vsearch to identify and remove chimeras:

vsearch -cluster_unoise unique_seqs.fa --centroids denoised.fasta

vsearch -uchime3_denovo denoised.fasta --sizein --sizeout --nonchimeras zotus_nochime.fa --relabel zotu

followed by this command to make an OTU table (table with each ASV/OTU's abundance in each sample):

vsearch -usearch_global filtered.fasta --db zotus_nochime.fa --id 0.97 --otutabout otu_counts_its.txt

Does this process preserve ASVs, or does the usearch_global step using --id 0.97 cluster the ASVs at 97% identity (essentially, convert them into OTUs)? When I do this, I get the same number of OTUs (labeled as zotus in my table, which I understand as being essentially the same as ASVs) as the number of non-chimeras listed in the output of the -uchime3 command,* which suggests to me that this process keeps the ASVs/zOTUs rather than making OTUs at 97% identity. But, if this is the case, what does the --id option in the command do?

*For example, in one of my datasets, running  vsearch -uchime3_denovo denoised.fasta --sizein --sizeout --nonchimeras zotus_nochime.fa --relabel zotu gives me the following output:

Found 1982 (16.6%) chimeras, 9975 (83.4%) non-chimeras,
and 0 (0.0%) borderline sequences in 11957 unique sequences.
Taking abundance information into account, this corresponds to
197868 (0.1%) chimeras, 162711937 (99.9%) non-chimeras,
and 0 (0.0%) borderline sequences in 162909805 total sequences.

9975 is the number of zOTUs/OTUs(?) I have in the otu_counts_its.txt table after running the usearch_global command

**Prior to these steps, I have dome some quality filtering and a dereplicating step, but no clustering steps.

Thanks in advance!

[Colin J Brislawn]

Hello Maria,

You are correct! This process preserves the ASVs/ZOTUs because all it's doing it mapping / counting the reads in filtered.fasta against the ASVs in zotus_nochime.fa and returning the hits in a feature-table.

Your observation that the number of features matches the non-chimeras confirms this!

In this context, --id 0.97 acts as a mapping threshold, so reads are <97% similar to any ASV are discarded. I've seen people do this step at 100% so that sequencing errors are not even counted, but I like a lower threshold like 99% or 97% so more of my reads are included in the count table. None of these settings change the ASVs that are in zotus_nochime.fa or will be in the output tables.

Colin

PS. zOTUs are what Robert Edgar calls ASVs. ASVs are also Operational Taxonomic Units, without the taxonomic part!

[Marita White]

That makes sense, thanks! One follow-up question: If there are sequences that are 97% similar to more than one ASV, are they counted in each ASV, or are they only counted once?

[Colin J Brislawn]

Another good question!

By default, they are only counted once because usearch_global only returns one hit for each input read.

Vsearch can report multiple hits, but I'm not sure I would want double/triple/multi counting here... and I'm not sure if --otutabout even allows higher values of --maxaccepts and/or --maxhits but you can try it and see.

Colin

[Marita White]

Makes sense, thanks!