Setting filtering thresholds after taxonomic assignment

[annsantos (Ana Santos)]

Hi

When we do the taxonomic assignment what cut-off was used for absence/presence of a taxon in each group e.g. in what proportion of the group did the taxon need to be present to be considered present in the group?

Thank you

Ana

[Nicholas_Bokulich]

Hello again @annsantos,
I have adjusted the title of this topic to better describe your question (to make it more searchable by others with the same issue!)

@annsantos

When we do the taxonomic assignment what cut-off was used for absence/presence of a taxon in each group

The taxonomy classifiers do not take abundance into account when classifying... they classify whatever sequences are given as input.

@annsantos

in what proportion of the group did the taxon need to be present to be considered present in the group?

This is something that would be done up/downstream in your workflow, and is largely based on your own workflow decisions. Singletons/rare features are filtered by default by the denoising methods, depending on which method you used. You can also use the filter-features method to filter features based on abundance or prevalence thresholds.

So to clarify: these proportions are defined by you in separate actions. This is done so by design, to give you better control over the process.

[annsantos (Ana Santos)]

Hi,

May be i put the question wrong. What i want to know is that determines if an ASV is e.g streptococcus. OTUs are sequences clusters that differ by less than a dissimilarity threshold, normally 3%, that is, those OTUs sequences represent a certain taxonomic unit with 97% similarity. And ASV?

[Nicholas_Bokulich]

Hi @annsantos,
I recommend checking out the overview tutorial and the literature cited in there to get a better idea of the differences between these steps. It sounds like you are conflated a few distinct processes: denoising, abundance filtering (which is often included in the denoising pipelines as I mentioned above), and taxonomy classification.

@annsantos

What i want to know is that determines if an ASV is e.g streptococcus.

taxonomy classification. These do not take abundance into account.

@annsantos

OTUs are sequences clusters that differ by less than a dissimilarity threshold, normally 3%, that is, those OTUs sequences represent a certain taxonomic unit with 97% similarity. And ASV?

ASVs are unique sequence variants... see the overview tutorial and especially the original publications on those methods for a better description of what they do. You can think of ASVs as 100% OTUs that have had some quality filtering applied to remove/correct sequencing errors.