Open thermokarst opened 6 years ago
For reference, here was a related thread from the QIIME 1 era: https://github.com/biocore/qiime/issues/1183
CCA provides a lot more info than PCoA, including association with explanatory variables (see here), so would be a really cool method to support. AFAIK it's commonly used in soil/ecology research, though I don't know if there are issues with compositionality.
This recently came up on the forum.
This would be a good thing to support! The technical aspects in emperor need some work but it wouldn't bee too hard to make it happen if we coordinate some concerted effort.
It would be great to see CCA as well as other methods from vegan. I had discussions with vegan creator Jari Oksanen about adapting vegan functions for Python last week in Oulu. Presumably this would be part of q2-diversity rather than a future q2-vegan plugin. Also see issue #123.
Slightly unrelated: I pushed a PR to Emperor with the necessary functionality to display biplots https://github.com/biocore/emperor/pull/646
Jari suggested the main dependencies of the vegan ordination methods are (with their Python implementations):
numpy.linalg.svd
scipy.linalg.qr
So I think the main dependencies are in place. I'm cc'ing @mortonjt so he is in the loop.
Thanks @cuttlefishh! Porting these functions to Python would be great, but it is worth pointing out that it isn't a strict requirement for getting them exposed in QIIME 2 as methods. We have the ability to wrap non-python functionality (for examples, check out q2-dada2, q2-phylogeny, and q2-alignment, to name a few). The only real requirement for official QIIME 2 plugins is that the dependencies are available via conda
- but that is only for our official plugins - unofficial plugins are welcome to do whatever they want for dependencies.
Just curious if this was added as a plugin to qiime diversity ?
No, not yet - are you interested in contributing?
References This type of analysis would be an alternative to emperor-based PCoA plots, as I understand it.