eviiom, canonically pronounced "ev-ee-ome", constructs a BIOM table of evidence of feature associations from two input BIOM tables.
To construct the evidence matrix, eviiom evalutes each pair of observations between the two tables (the "left" and "right" tables). If both observations are found to exist in the same sample, evidence for an association is incremented (i.e., positive evidence). If a left observation is found without a right observation, under the constraint that a positive association had been found, evidence for the pair is decremented (i.e., contradictory evidence).
A formal setup.py is not available yet. To install:
$ git clone https://github.com/biocore/eviiom.git
$ cd eviiom
$ export PYTHONPATH=`pwd`:$PYTHONPATH
$ python eviiom/drive.py blend --help
Usage: drive.py blend [OPTIONS]
Options:
--left-table PATH A BIOM table [required]
--right-table PATH A BIOM table [required]
--threads INTEGER
--min-overlap INTEGER The minimum number of supporting samples overlapping
--output PATH [required]
--left-as-observations Set the observation IDs in the left table as the
observation IDs in the resulting table
--help Show this message and exit.
eviiom requires numpy, click, biom >= 2.1.6 (with h5py), and joblib. Version requirements noted where known.
The samples axis must be in common between both tables in terms of the IDs represented (order does not matter). The two input tables are described as "left" and "right." The values within the "left" table are used to increment and decrement evidence. The observation axis of the output table can be controlled at runtime so it is relative to either the left or the right tables.