Predicting abundance but requiring abundance

[freiburgermsu]

I'm confused why MICOM returns the arbitrary initial abundances in the grow results. I thought that MICOM predicts the abundances of a given community; is this not true?

[cdiener]

No this is not true. In a steady state model abundances are fixed and to a certain degree arbitrary (the docs have a section on the methods but its a bit out of date). MICOM predicts growth rates and fluxes.

[freiburgermsu]

Aren't growth rates the same as abundances when controlling for time, such as ( member coculture growth rate / community growth rate ) ? or maybe abundance is more similar to ( (member coculture growth rate * abundance) / community growth rate) ?

[cdiener]

You mean they are correlated in steady state? That depends on the specific model, environment and assumptions you make. Under arbitrary dilution rates and death processes they are fairly unrelated. MICOM does push the growth rates towards a solution that is correlated with the abundances in rich media. However, taxa may always violate this. The simplest example is a taxon missing an essential cofactor in the medium. This taxon will always receive a growth rate of zero even if the abundance has been set to a non-zero abundance. This would still be a valid steady state under zero dilution.

If you make assumption about the dilution process you might get different relationships. For instance a constant mass-action dilution ends up in the SteadyCom approach where each each taxon has the same fixed growth rate.

[freiburgermsu]

Thank you for the elaboration on growth rates and abundances. Are you saying that the taxon with a missing cofactor will always in FBA have a 0 growth rate, or that in a natural community it will? I guess I don't understand how it can have any abundance in the steady-state if it does not have any growth?

Separately, why the does the regularization exist in MICOM if the abundances are fixed? I thought that the regularization minimized variance among the member abundances, but if the abundances are fixed this can't be the point of regularization?

[cdiener]

Are you saying that the taxon with a missing cofactor will always in FBA have a 0 growth rate, or that in a natural community it will?

Mostly in FBA. If it is detectable in nature I would assume it has a way of getting the cofactor.

I guess I don't understand how it can have any abundance in the steady-state if it does not have any growth?

I agree it should not, the motivation for building MICOM was that present taxa should have non-zero growth rates. Just with the steady state assumption this is possible though. In that case the dilution/death rate would be zero. In practice we never consider the steady state infinetly long, just somewhat long like a full exponential growth phase. An analogy would be you swallowing a bubble gum. It will be abundant in your gut but not grow. Of course if will be pooped out at some point but some longer period it will have a non-zero abundance but a zero growth rate.

I thought that the regularization minimized variance among the member abundances, but if the abundances are fixed this can't be the point of regularization?

It regularizes the growth rates, not the abundances. You are correct in that abundances are fixed in a MICOM model.