Background: The accumulation of knowledge on biological networks and high-throughput experimental data raises the need of robust, efficient, schematic and extendable method on network dynamical modeling and control analysis.Although there are some algorithms and software available, efficient method that satisfies all the requirements is still lacking.Methods: We developed a modified simulated annealing(MSA) algorithm based on local parameter sensitivity of the objective function, which can record dependence of metabolite integral area on the kinetic parameters simultaneously, thus providing guidance for key targets identification by searching for targets that affect production of metabolites concerned without much influences on other metabolites.This framework has been applied in fitting metabolite concentration data in the rat blood arachidonic acid metabolic network from LCMS experiments.Results: We found 10 different sets of kinetic parameters and initial concentrations using the modified simulated annealing algorithm.Simulation with these sets of parameter and initial concentration proves consistence with experimental data.In the simultaneous target identification process, we found seven candidate anti-inflammatory targets including COX2, PGES, 5-LOX and LTA4H, whose inhibitors are proved to be effective in inflammation intervention.The modified simulated annealing algorithm shows faster convergence to better solutions compared to the original simulated annealing algorithm and differential evolution(DE) in parameter estimation of arachidonic acid metabolic network in human platelets.Conclusions: Combination of modified simulated annealing and selective target identification can increase the efficiency of calculation and provide more accurate information on disease related molecular network intervention.This methodology can also be used in modeling and controlling of other biochemical networks .