CoRunner: multiple optimization run manager for Copasi software
2012
Sulins, J., TIBIT, Ltd., Ozolnieki, Ozolnieku pag., Ozolnieku nov. (Latvia) | Stalidzans, E., Latvia Univ. of Agriculture, Jelgava (Latvia). Faculty of Information Technology. Dept. of Computer Systems. Biosystems Group
Numerical optimization based design tasks of dynamic models of biochemical networks mainly are solved using global stochastic optimization methods which can deliver relatively quick improvement of the objective function in short time but it cannot guarantee finding of optimal solution. In case of large models or complicated dynamics the question about appropriate duration of optimization runs become important question in case of necessity to analyze high number of combinations of adjustable parameters. This task becomes even more complex if the optimal solution has to fulfil the steady state condition which is typical in case of biotechnological applications of design tasks as steady state precondition rejects many combinations of adjustable parameter values thus making the optimization slower. Duration of one optimization run therefore can take time up to several days or weeks for one combination of adjustable parameters which make the procedure inefficient. Application of parallel optimization runs is proposed to reduce the stochastic impact of a single optimization run. A software CoRunner is developed to apply parallel runs automatically collaborating with software COPASI for automation of optimization procedure. In case of necessity the dynamics of parallel optimization runs, which are stored in form of COPASI reports, can be visualized with software ConvAn for detailed analysis of optimization performance. CoRunner automatically stops the optimization in case of consensus of all the parallel optimization runs and returns COPASI model with the set of best values of adjustable parameters found.
显示更多 [+] 显示较少 [-]