How to set media to interpret and compare the metabolic interactions at different habitats?

[CODRbio]

What are you trying to achieve? What question are you trying to answer?

Thank you for providing the pipeline that allows us to understand the interactions of microbes in different habitats. I would like to compare the potential levels of metabolic interactions in deep-sea habitats. I have hypothesized that microbial interspecies interaction potential (MIP) should be stronger in habitats with imbalanced nutrient input, such as the seepage of cold seeps. However, after reading the issues in metaGEM and Smetana, I'm still unsure how to proceed effectively.

What microbial community are you studying? e.g., Human gut, free-living soil, synthetic lab culture?

I'm studying natural sediments and water.

Detailed question.

As proposed in the Smetana issues, I should use --flavor bigg to get the MIP. However, in the metaGEM snakefile, I found you are using --flavor fbc2. Is this acceptable for the global mode?

The main problem is getting the details of the interactions. After reading your responses to the issues, I think my failure to get detailed interactions might be due to using the same media for gapping and Smetana. However, when I use --flavor bigg, the global mode provides values for global MIP but nothing for detailed Smetana. According to your suggestions, when I do gapping and Smetana with the same media, the MIP should be NA. How can I get the global MIP?

As I'm working with MAGs, some pathways are incomplete. So, if I want to understand the interactions within the community, should I first gapfill the model with the full media and then use Smetana with my own media?

[CODRbio]

I've tried to gap model with the full media and smetana the community with the minimum media. The global mode give nothing with the warning UserWarning: MRO: Failed to find a valid solution for community
warn('MRO: Failed to find a valid solution for community')
~

[CODRbio]

Also cannot find solution in the detailed mode.
/nfs_genome/anaconda/envs/metaGEM_cplex/lib/python3.7/site-packages/smetana/smetana.py:104: UserWarning: SCS: Failed to find a solution for growth of SAR324_RS_SUR_bin.10
warn('SCS: Failed to find a solution for growth of ' + org_id)
/nfs_genome/anaconda/envs/metaGEM_cplex/lib/python3.7/site-packages/smetana/smetana.py:104: UserWarning: SCS: Failed to find a solution for growth of HIMB59_NC_SUR_bin.53
warn('SCS: Failed to find a solution for growth of ' + org_id)
/nfs_genome/anaconda/envs/metaGEM_cplex/lib/python3.7/site-packages/smetana/smetana.py:104: UserWarning: SCS: Failed to find a solution for growth of Nitrososphaerales_NC_SUR_bin.118
warn('SCS: Failed to find a solution for growth of ' + org_id)
/nfs_genome/anaconda/envs/metaGEM_cplex/lib/python3.7/site-packages/smetana/smetana.py:104: UserWarning: SCS: Failed to find a solution for growth of HIMB59_SEEP-bub50_bin.12
warn('SCS: Failed to find a solution for growth of ' + org_id)

[CODRbio]

I've read most of your reply for metaGEM, so I need to try to use rich media for gapping and poor media for stimulation, am I right? But I need to make the media work for most species, so must I adjust the media myself?

[franciscozorrilla]

Hi @CODRbio,

Here are some comments:

1. As proposed in the Smetana issues, I should use --flavor bigg to get the MIP. However, in the metaGEM snakefile, I found you are using --flavor fbc2. Is this acceptable for the global mode?

The Snakefile has the SMETANA detailed mode implementation (-d flag), not the global metrics (-g). Based on comments from issues in the SMETANA repo, it looks like you can try changing the flavor flag in case you are not getting results with your models (cdanielmachado/smetana#22), although it might not necessarily solve your issues.

2. The main problem is getting the details of the interactions. After reading your responses to the issues, I think my failure to get detailed interactions might be due to using the same media for gapping and Smetana. However, when I use --flavor bigg, the global mode provides values for global MIP but nothing for detailed Smetana. According to your suggestions, when I do gapping and Smetana with the same media, the MIP should be NA. How can I get the global MIP?

I do not think that I have suggested what is italicized in the comments above, can you link me to that suggestion? From my presonal experience, I have been using SMETANA mainly for analyzing possible detailed interactions, not so much the global metrics. I would suggest playing around with the flavor flag using a small toy community to see how it affects the reported results. Check out this paper for more on global metrics usage and interpretation.

3. As I'm working with MAGs, some pathways are incomplete. So, if I want to understand the interactions within the community, should I first gapfill the model with the full media and then use Smetana with my own media?

Just to clarify, were your MAG-based GEMs reconstructed using CarveMe or some other tool? As noted in the linked issue (cdanielmachado/smetana#22), SMETANA can run into problems with non-CarveMe GEMs due to formatting differences. Regarding gapfilling, there is a tradeoff between proprely fixing incomplete pathways and incorrectly resolving true auxotrophies that may occur in the microbes, unfortunately since many of them are not culturuable and well studied it is difficult to determine. If I were you, I would try the analysis both with gapfilled and then with non-gapfilled GEMs, you should expect to see more exchanges in non-gapfilled.

4. I've read most of your reply for metaGEM, so I need to try to use rich media for gapping and poor media for stimulation, am I right? But I need to make the media work for most species, so must I adjust the media myself?

Thats right, you gapfill on rich media to avoid overgapfilling, then simulate detailed interactions in a poorer media in order to see how microbial interactions can support community members. I do indeed recommend playing around with the media formulation to see how it may affect the gapfilling and simulation. For example, check out this recent cheese study where we applied a similar analysis.

Best wishes,
Francisco

[CODRbio]

Thank you for your quick recommendations. I'll be incorporating your suggestions and feedback as I work towards my goals.

I do not think that I have suggested what is italicized in the comments above, can you link me to that suggestion?

I'm sorry for the misunderstanding of your suggestions for the detailed stimulations. After reading more information, I realized that there are difference between the two modes.

Just to clarify, were your MAG-based GEMs reconstructed using CarveMe or some other tool? there is a tradeoff between proprely fixing incomplete pathways and incorrectly resolving true auxotrophies that may occur in the microbes, unfortunately since many of them are not culturuable and well studied it is difficult to determine.

I get the models with the carveme. I agree with you, so I'd like to do a series of dry experiments to see the influence and try to get a reasonable explanation for the consensus results.