search

opencobra / memote

memote – the genome-scale metabolic model test suite
https://memote.readthedocs.io/
Apache License 2.0
128 stars 28 forks source link

Same file but different consistency result between versions #708

Closed BT287 closed 3 years ago

BT287 commented 3 years ago

Problem description

Hi, I was trying to assess my model but got a problem in testing consistency parts. When I assessed my model with memote of latest version (0.12.0), it returns stoichiometric consistency as false. But it returns 99.9% when the same model is assessed by memote of version 0.11.1, which is provided by memote website (https://memote.io/).

I want to find out what makes my model got "false" in the stoichiometric consistency result with the latest version.

Here is the status of my virtualenv,

System Information

OS Linux OS-release 4.4.0-193-generic Python 3.6.7

Package Versions

Jinja2 2.11.2 black ; extra == 'development' not installed click 7.1.2 click-configfile 0.2.3 click-log 0.3.2 cobra (>=0.15.3) not installed cookiecutter 1.7.2 depinfo 1.6.0 future 0.18.2 gitpython 3.1.12 goodtables (~=2.0) not installed importlib-resources 5.0.0 isort ; extra == 'development' not installed lxml 4.6.2 memote 0.12.0 numpydoc 1.1.0 pandas 1.1.5 pip 20.3.3 pylru 1.2.0 pytest (>=4.0) not installed requests 2.25.1 ruamel.yaml (>=0.15) not installed setuptools 51.1.2 six 1.15.0 sqlalchemy 1.3.22 sympy 1.7.1 tox ; extra == 'development' not installed travis-encrypt (<1.2.0) not installed wheel 0.36.2

model_file_and_result.zip

carrascomj commented 3 years ago

The results from both versions are the same but the metric reported is different. In version 0.12.0, the stoichiometric consistency is a boolean metric. Having that the weight on the stoichiometry consistency is 3, 53.39% seems fine to me.

In version 0.11, it seems like it was reporting 1 - (Unconserved Metabolites / Total number of metabolites). Now, the unconserved metabolites and the minimal net stoichiometries are displayed as information (they can be found in the right column) and do not influence the score, but they report the same results for your model in both versions.

BT287 commented 3 years ago

I see. I didn't know that stoichiometric consistency changed to a boolean metric. Thanks for letting me know that. Good luck!