denoise-*: expose `--collapse-no-mismatch` parameter

[nbokulich]

References
forum xref

[jakereps]

I can take this one 👍

[jakereps]

How would we want to implement this? That doesn't appear to be a parameter, but a standalone function within the dada2 package.

collapseNoMismatch <- function(seqtab, minOverlap=20, orderBy="abundance", vec=TRUE, verbose=FALSE)

By expose is it adding a new method to q2-dada2, or having this flag run it using strictly the default parameter values?

[ebolyen]

I suspect the latter is what we're looking for, since it sounds like the forum user was interested in combining subsequences together. I'm a little curious as to what minOverlap is for, since it seems like we're dealing with +- a few nucleotides, so having a subsequence of only 20ish sounds pretty bad (and will probably bin into the most abundant read no matter what).

[ebolyen]

That being said, a new method would let us re-use this elsewhere...

[nbokulich]

that forum post seemed to suggest that this was a parameter, so that was my assumption when I opened this issue without further investigation.

I like the new method idea though.

[jakereps]

I'm not sure the best way to handle adding the function inline w/ denoise, as it'd require the flag to run it or not and also a flag for the overlap parameter - assuming all the other default values are fine to use. I'll start with just the function I suppose. Would joining them have to be a pipeline then? It could always be added inline to denoise as well and keep this as a separate function. ¯\_(ツ)_/¯

[thermokarst]

I'm not sure the best way to handle adding the function inline w/ denoise, as it'd require the flag to run it or not and also a flag for the overlap parameter

I think this would be conditionally run in the R scripts, right (like the chimera checking section)?

[benjjneb]

To "qiime" in, if this is something that is going to be integrated into the workflow commands, this may be best left until we migrate the plugin to the version 1.8 of the dada2 R package, as I am expecting to overhaul the q2-dada2 R scripts a bit at that time.

p.s. we are waiting on bioconda updates for that migration to happen: #85 (comment)

[fconstancias]

If I am not wrong, the 100% clustering of ASV can also be performed using q2-vsearch plugin ? So maybe there is no need to add a specific condition in q2-dada2 script?
What do you think?

[ebolyen]

I'm a bit uncertain if our implementation of de-novo will cluster shorter sequences or not. Also, the IDs are something to consider, but I don't see why these couldn't be made to be very compatible.