Journal of Biological Systems, Vol. 22, No. 1 (2014) 101–121 c  World Scientific Publishing Company DOI: 10.1142/S0218339014500065 USING AN IMPROVED BEE MEMORY DIFFERENTIAL EVOLUTION ALGORITHM FOR PARAMETER ESTIMATION TO SIMULATE BIOCHEMICAL PATHWAYS CHUII KHIM CHONG ∗,‡ , MOHD SABERI MOHAMAD ∗,§ , SAFAAI DERIS ∗,¶ , MOHD SHAHIR SHAMSIR †,‖ , LIAN EN CHAI ∗,∗∗ and YEE WEN CHOON ∗,†† ∗ Artificial Intelligence and Bioinformatics Research Group Faculty of Computing † Department of Biological Sciences Faculty of Biosciences and Medical Engineering Universiti Teknologi Malaysia 81310 UTM Skudai, Johor, Malaysia ‡ ckchong2@live.utm.my § saberi@utm.my ¶ safaai@utm.my ‖ shahir@fbb.utm.my ∗∗ lechai2@live.utm.my †† ywchoon2@live.utm.my Received 3 April 2013 Accepted 15 July 2013 Published 17 January 2014 When analyzing a metabolic pathway in a mathematical model, it is important that the essential parameters are estimated correctly. However, this process often faces few prob- lems like when the number of unknown parameters increase, trapping of data in the local minima, repeated exposure to bad results during the search process and occurrence of noisy data. Thus, this paper intends to present an improved bee memory differential evo- lution (IBMDE) algorithm to solve the mentioned problems. This is a hybrid algorithm that combines the differential evolution (DE) algorithm, the Kalman filter, artificial bee colony (ABC) algorithm, and a memory feature. The aspartate and threonine biosynthe- sis pathway, and cell cycle pathway are the metabolic pathways used in this paper. For three production simulation pathways, the IBMDE managed to robustly produce the estimated optimal kinetic parameter values with significantly reduced errors. Besides, it also demonstrated faster convergence time compared to the Nelder–Mead (NM), sim- ulated annealing (SA), the genetic algorithm (GA) and DE, respectively. Most impor- tantly, the kinetic parameters that were generated by the IBMDE have improved the production rates of desired metabolites better than other estimation algorithms. Mean- while, the results proved that the IBMDE is a reliable estimation algorithm. Keywords : Parameter Estimation; Differential Evolution Algorithm; Kalman Filter; Artificial Bee Colony Algorithm; Memory Feature. § Corresponding author. 101 J. Biol. Syst. 2014.22:101-121. Downloaded from www.worldscientific.com by UNIVERSITI TEKNOLOGI MALAYSIA on 08/12/14. For personal use only.