Metadata state-column can't have a zero value or be non-numeric category

[mestaki]

Hey @cameronmartino,
Finally getting around to trying out this awesome tool, huge congrats on the paper btw.

I was trying a dataset with paired-data with just 2 timepoints when I ran into a problem with my metadata state-column.
I tried 2 different columns that could have represented my states:
timepoint <- has numeric values of 0 and 6 which correspond to weeks, or
Est_status <- categorical values of Baseline, and 6mosEst

Both represent the same values one is numeric the other categorical. I'm on q2-2020.6, with q2-deicode: 0.2.4, and gemelli: 0.0.6 installed
When I run the below with either of those 2 columns:

!qiime gemelli ctf \
    --i-table table.qza \
    --m-sample-metadata-file pilot_metadata2.txt \
    --m-feature-metadata-file taxonomy_silva132.qza \
    --p-state-column Est_status \
    --p-individual-id-column Patient_ID \
    --output-dir gemelli/ \
    --verbose

I get the following error:

/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/numpy/core/fromnumeric.py:3335: RuntimeWarning: Mean of empty slice.
  out=out, **kwargs)
Traceback (most recent call last):
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/q2cli/commands.py", line 329, in __call__
    results = action(**arguments)
  File "<decorator-gen-555>", line 2, in ctf
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/qiime2/sdk/action.py", line 245, in bound_callable
    output_types, provenance)
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/qiime2/sdk/action.py", line 390, in _callable_executor_
    output_views = self._callable(**view_args)
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/gemelli/ctf.py", line 40, in ctf
    feature_metadata)
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/gemelli/ctf.py", line 129, in ctf_helper
    n_initializations=n_initializations).fit(tensor_rclr(tensor.counts))
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/gemelli/factorization.py", line 171, in fit
    self._fit()
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/gemelli/factorization.py", line 214, in _fit
    fillna=self.fillna)
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/gemelli/factorization.py", line 534, in tenals
    if rank_estimate(obs_tmp, eps_tmp) >= (min(obs_tmp.shape) - 1):
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/gemelli/optspace.py", line 462, in rank_estimate
    r_one = np.argmin(cost)
  File "<__array_function__ internals>", line 6, in argmin
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/numpy/core/fromnumeric.py", line 1267, in argmin
    return _wrapfunc(a, 'argmin', axis=axis, out=out)
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/numpy/core/fromnumeric.py", line 58, in _wrapfunc
    return _wrapit(obj, method, *args, **kwds)
  File "/home/mestaki/miniconda3/envs/qiime2-2020.6/lib/python3.6/site-packages/numpy/core/fromnumeric.py", line 47, in _wrapit
    result = getattr(asarray(obj), method)(*args, **kwds)
ValueError: attempt to get argmin of an empty sequence

Plugin error from gemelli:

  attempt to get argmin of an empty sequence

See above for debug info.

If I add a random 3rd value in either of these columns it runs successfully. So can gammeli not be used on data with only 2 timepoints then? It doesn't fit in the cross-sectional version either, so this would be a big loss since many projects do use only before/after designs.

[cameronmartino]

Thanks for reporting and the detailed issue @mestaki! This is a bug and CTF should run for two subject/state tensors. I have put in a specific issue and PR to fix this bug.

[mestaki]

Howdy, just wondering if there are any updates on this fix, or estimated timeline for a fix? I want to gauge if I should wait it out for a project I'm working on or try a different approach (fingers crossed for the former).

[cameronmartino]

Hi @mestaki

Yes, there is a branch with a bug fix waiting for a PR review. You can install and use that branch by running pip install git+https://github.com/biocore/gemelli@two-time-bug.

Thanks for the reminder!

[mestaki]

Fantastic, thank you!