Closed-form poisson differential computation.
All of the code in poisson_cat.py was written by Jamie Morton
(@mortonjt)
and Cameron Martino (@cameronmartino),
who also worked out the accompanying math. See
here for
the original gist from Jamie, and
here
for Cameron's version of it.
pip install click biom-format pandas numpy scipy scikit-learnUsage: run.py [OPTIONS]
Options:
-t, --table TEXT BIOM table with count data [required]
-m, --metadata TEXT Sample metadata file [required]
-c, --category TEXT Metadata category of interest; currently
only binary categories (i.e. those
containing only two unique values) are
supported [required]
-r, --reference-category TEXT Reference metadata category of interest; if
not specified, the first category will be
picked
-o, --output-path TEXT Output filepath to which differentials TSV
will be written [required]
-f, --filter-category-value TEXT
If passed, this will filter out all samples
with a -c category value of this string.
This will also afterwards filter out all
"empty" features. This is useful if you have
a category with three possible values that
you'd like to make into a binary category,
so that it can be used here.
--help Show this message and exit.
The output differentials can be displayed in
rankratioviz. If you're using
rankratioviz through QIIME 2, you can import the output differentials here as a
FeatureData[Differential] artifact, and visualize them using
qiime rankratioviz supervised-rank-plot; if you're using rankratioviz
outside of QIIME 2, you should just be able to pass the differentials directly
to rankratioviz.
(Note that you might want to use the -x / --extreme-feature-count option in
rankratioviz if your BIOM table has a lot of entries.)