
Represent complex fermentative, transport etc. process as a system of
elementary ( first order) reactions, e.g. associations/dissociations, and
transformations. In such a way reactions like a calcium uniport,
respiratory complex I functioning, glutathione peroxidase work could be
modeled. Submodel approximation level depends on the it's role in the whole
model system  more important system part could be modeled in more details. For example, 2substrate 1product reaction will look in the following way: This representation is unified for all reactions and convenient for estimation in matrixhandling programs, like MatLab.  
Assign reaction constants and substances amounts (from the experimental data or by model fitting)  
(An alternative approach, which is in development now: representing reactions as complex processes with predefined kinetics, e.g. substrate inhibition, reversible Hill's kinetics etc. It is more convenient, but less precise approach.)  
Submodel optimization according to experimental data  fitting to experimentally defined system parameters' values, dependencies and change with time.  
Submodel behavior analysis  stability range and states analysis, bifurcation points, oscillation regimes etc. 
Small submodels are compiled into bigger ones. In this way
severalreactions models are formed, e.g. calcium transport system (i.e.
system consisting of calcium uniport & Ca/Na exchange & Ca buffering
in mitochondrial matrix), respiratory chain model, mitochondrial
freeradical defense system (GSHperoxidase & GSSGreductase &
GSHtransferase & PNtranshydrogenase & PNdependent dehydrogenases
etc.). Below you can see submodel of main processes of mitochondrial freeradical defense system:  
Growing models should be optimized (in accordance to experimental data)
after each small submodels addition. Here is general scheme fragment including parts of freeradical defense, Pitransport and calcium transport systems:  
Severalreactions models, after optimization, are compiled into the general model of the
process:  
General model optimization according to experimental data.  
"Numerical experiments" with the general model: Possibly, big complex models should not be completely analyzed in a usual way (complete stability and states analysis, bifurcation points finding etc.) due to huge number of components. We can use rather "numerical experiment", when we observe behavior of the given complex system in prescribed conditions, and are able to change any parameter in some range (defined from submodels optimization and analysis, according to the real system) to change system behavior and define control coefficients. 