Archaea reconstruction does not predict Methanogenesis anymore

[cdiener]

Hi,

after some issues with Archaea reconstructions not predicting methanogenesis we tried to debug with the example provided in the docs. The exact assembly used in the example has been removed from Refseq but using the new one for the same organism will not yield methanogenesis anymore when running the exact sequence of commands from here on the current master branch of gapseq.

[jotech]

hi @cdiener
Thank you for reporting the issue!

I looked into it, and the methanogenesis predictions still work for M. barkeri (GCF_000195895.1), M. smithi (GCF_000016525.1), and also M. thermautotrophicus (GCF_000008645.1).

It seems that alternative solutions are causing the issue since methane production is present for all organisms with fba but only for M. smithii with mtf/pfba.

I constructed the models with the latest gapseq version (1.3.1) and used the predicted medium.

I will also check the new genome and the minimal medium from the tutorial, but my best bet is it's related to alternative solutions.

[cdiener]

Hi @jotech,

Thanks for looking into this. Was the METHANOGENESIS-PWY found for all of those in the initial reaction search? Or did it get added in during gapfilling? For us, the core issue is that it is not found anymore in the initial reaction step, even for organisms that should have the pathway. Looking at the reaction tables it seems that it covers the same reactions as in previous versions but the subunits for two enzymes have changed which leads to those enzymes not being found anymore. It does work for us if we lower the bitscore cutoff to 100.

[jotech]

That's a valid point; I did not consider it; thanks!

I have a pathway (METHANOGENESIS-PWY) found for M. barkeri (83% completeness) and partially also for M. smithii (66% completeness) using the default bitscore. But I agree, this should have been more evident. I will have a closer look. Thank you for the hint with the subunits!