The reproducibility issues of SCFAs prediction results using micom

[LiangenLuo]

Hi,

I find that every time I run 'qiime micom grow' or 'micom.workflows.grow', the resulting 'fluxes' are different, leading to varying SCFAs production for each sample.
I would like to know how to resolve this issue.

Thanks

[cdiener]

Hi. What are you using as strategy for the grow function? And which solver, database and medium? Running the same samples without changes should usually give consistent results even though fluxes don't have to be unique (algorithm is deterministic).

[LiangenLuo]

I'm sorry, I didn't describe it clearly. I'm using the sample data and example code from the q2-micom Community Tutorial.
solver: qiime2-amplicon-2024.2-py38-linux-conda.yml
Here is my code:
qiime micom build --i-abundance cdi_table.qza
--i-taxonomy cdi_taxa.qza
--i-models agora103_refseq216_genus_1.qza
--p-cutoff 0.0001
--p-threads 40
--o-community-models models.qza
--verbose
qiime micom grow --i-models models.qza
--i-medium western_diet_gut_agora.qza
--p-tradeoff 0.5
--p-threads 40
--o-results growth.qza
--verbose

Problems found:
I ran the grow model twice without changing anything, and I noticed that the results in the file 'exchange_fluxes.csv' showed different 'flux' values. Then I ran it a third time and found that the results were inconsistent with the previous two runs. such as：

running taxon sample_id tolerance reaction flux abundance metabolite direction
1 Fusobacterium ERR1883240 1.00E-06 EX_but(e) -28.24858731 0.095046625 but[e] import
2 Fusobacterium ERR1883240 1.00E-06 EX_but(e) -12.04266508 0.095046625 but[e] import
3 Fusobacterium ERR1883240 1.00E-06 EX_but(e) -19.25519755 0.095046625 but[e] import

On the other hand, the total number of output results is also different：
running results (rows)
1 16104
2 16142
3 16119

Finally, it was found that the production of SCFAs obtained were different. such as:
running sample_id but[e] ppa[e]
1 ERR1883207 6.019178823109127 2.1169882770011683
2 ERR1883207 6.973371470007656 2.1169882770011683
2 ERR1883207 6.298594045844579 2.116988277001168