Forecast of numerical optimization progress of biochemical networks

Increasing size of biochemical models of cellular processes influence the time consumption for optimization of control systems in biotechnological plants to increase the productivity. Nonlinear systems of differential equations can be optimized by time consuming stochastic numerical methods that work for most different nonlinear models but can not guarantee finding of the global optimum. In case of a high number of possible parameter combinations early rejection of parameter combinations with low potential of criteria increase becomes important. Statistics of convergence dynamics is collected to predict the optimization potential of optimization parameter combinations and use them for early rejection of parameter combinations with low optimization potential. A prediction tool is developed to predict the distance to the global optimum depending on the number of parameters, size of the model, and the number of parameters in the model. The prediction tool returns distance to the expected optimal solution as a function of Central Processor Unit (CPU) time. The forecast tool can be used for models of different size and using different numerical optimization methods.