Comparison of parameter fitting on the model of irradiation effects on bystander cells between Nelder-Mead simplex and particle swarm optimization

Abstract

Study on the biological effects of irradiation has become important nowadays. Mathematical modeling is one of the interests among researchers due to its ability to explain the dynamics process of the irradiation. Some physical parameters cannot be evaluated from the empirical data. Therefore, the aim of this work is to estimate parameters of the model of irradiation effects on bystander cells using optimization approaches. We employ two algorithms: Nelder-Mead Simplex (NMS) (which is the local optimizer) and Particle Swarm (which is the global optimizer). We compare the efficiency of two optimization algorithms in optimizing the parameter values of the model. 50 sets of parameters have been estimated and all sets are able to match the model simulation and the experimental data with the least Sum-Squared Error (SSE). The graph of model simulation using a set of the estimated parameters from both optimization algorithms shows a good fit with the experimental data. The overall results indicate that NSM is better than Particle Swarm (PS) optimization in the aspect of time computing, while there is no significant difference in the score of SSE and converging iteration to the least SSE.