Move table input away from BIOM

biocore / BIRDMAn

Bayesian Inferential Regression for Differential Microbiome Analysis

BSD 3-Clause "New" or "Revised" License

22 stars 5 forks source link

Move table input away from BIOM #6

Closed gibsramen closed 2 years ago

gibsramen commented 3 years ago

Brought up by @gwarmstrong.

The BIRDMAn input table does not need to be in BIOM and in fact is likely better off being switched to e.g. a simple NumPy matrix. This would have the added benefit of removing BIOM as a dependency, easing some of the weird installation issues. The future QIIME 2 implementation should hook into the input function with a qza -> BIOM -> NumPy transformation.

mortonjt commented 3 years ago

I agree with this -- but it can be an optional dependency for the qiime2 integration (where qiime2 is also optional).

There maybe scenarios where having the BIOM structure could be super useful to fully exploit sparsity; so in the distant future, it may be worthwhile to consider interoperability with csr / csc / coo formats.

gwarmstrong commented 3 years ago

it may be worthwhile to consider interoperability with csr / csc / coo formats.