FisMat2017 - Submission - View

Abstract's title: Statistical mechanics of metabolic networks
Submitting author: Daniele De Martino
Affiliation: IST Austria
Affiliation Address: AM Campus 1, 3400, Klosterneuburg, Austria
Country: Austria
Oral presentation/Poster (Author's request): Oral presentation
Other authors and affiliations:

How much is the metabolic state of bacteria optimal for growth (even in ideal conditions)?
Maximization is an usual assumption in modeling (FBA, flux balance analysis) but e.g. single cell growth rates fluctuate.
Inspired by statistical mechanics, here distributions in stationary metabolic networks at fixed average growth
(MaxEnt) are considered, with one parameter that interpolates from a random state to the FBA solution,
The parameter (selection strength, akin to inverse temperature) inferred from experimental flux estimates of the catabolic core of E.Coli
in several conditions is close to, but not at the optimality assumed by FBA.
This framework makes further predictions on flux variability, correlations and scaling,   
the latter being verified by single-cell growth rate data at different sub-inhibitory antibiotic concentrations.
Further, MaxEnt are the stationary state of logistic growth dynamics (used to fit OD curves) in the heterogeneous case,
revealing insights on inoculum size dependence.
Why not optimal? MaxEnt hints at "minimum information costs" (noise to reduce, e.g. in expression)
to achieve a certain optimization and shows (with further mild assumptions) a tradeoff between optimization
and response time (e.g. to upshifts) in the form of a fluctuation theorem as revealed by a map between coarse grained
equilibrium dynamics and the states of a quantum Airy particle subjected to a centrifugal force.   
If not growth, what is the "objective function" (if any) of cell metabolism in given conditions?
An inverse modelling  framework will be discussed where the latter can be directly inferred
from data and  some preliminary results on E. Coli will be shown. I will conclude with some computational remarks
on inference errors leading to scale free tails in flux distributions and  discuss
in which conditions phenotypic bistability arises  and can be described within a Van-Der-Waals picture.