Constrained Stochastic Space Search Method for Parameter Estimation in Biological Networks

Parameter estimation is an important part of computational systems biology – especially in studies on biological networks. Numerous stochastic search methods have been applied in parameter estimation in biological networks. In this paper, a constrained stochastic space search (CSSS) method for parameter estimation is proposed and evaluated for estimating the parameters of a genetic network described by differential equations. Both linear and nonlinear model formalisms were used for the data evaluation. The performance of the CSSS method was compared to the Integrated Controlled Random Search for Dynamic Systems (ICRS/DS) stochastic optimization algorithm. Compared to the ICRS/DS, the CSSS algorithm is faster with at least a 7-fold shorter convergence time. Independent replicates were run and identification performed. For the same initialization conditions prior to optimization, the CSSS had on average smaller relative mean errors than the ICRS/DS.