Flux Balance Analysis: Solver status is infeasible

[Kaushik Karambelkar]

 

Hello,


The model I'm working on is a pretty big model, about ~8000 reactions, ~4000 genes and ~5000 metabolites.

model.optimize() returns the following:-
<Solution infeasible at 0x7fcb2f415b50>

What can the problem be?

Thank You.

Kaushik

[Joshua Lerman]

Usually with these models the 0 vector is feasible, since with S*v = 0 if v is 0 the constraint is satisfied. Recall that S is the stoichiometric matrix.

I'd start by looking at any non 0 or 1000 or -1000 lower and upper bounds that you have on the model. What flux are you requiring such that the model is infeasible ?

Josh

--
You received this message because you are subscribed to the Google Groups "cobra pie" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cobra-pie+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Christian Diener]

Hi what version of cobrapy are we talking about? 0.5.11 or 0.6.x? 

In general infeasible means there is no solution that can fulfill all of the constraints. Sometimes it can also arise when optimizing unbounded functions but that is not common.

Usually it happens when you set a constraint that can not be fulfilled. The two most common cases I have seen are:

- forcing an import or demand reaction to have a non-zero flux without having a consumed reaction (metabolite can only accumulate so there is no steady state and S*v = 0 is violated)

- forcing a minimal value for the objective that can not be obtained (can happen when not taking the solver accuracy into account)

Especially the first one is often observed when using models generated by some upstream algorithm. Sometimes those add import or demand reactions for metabolites contained in the objective function. Thus maximizing the objective forces leaving steady state.