PDF file
Transcription
PDF file
Module Decomposition and Integration Method Optimizes A Large-Scale Cell Cycle Model Daichi Nitta1 Hiroyuki Kurata1 [email protected] [email protected] 1 Department of Bioscience and Bioinformatics, Kyushu Institute of Technology, Iizuka, Fukuoka 820-8502, Japan Keywords: CADLIVE, dynamic simulation, systems biology, budding yeast, cell cycle, GA 1 Introduction Major objectives of systems biology are to build molecular interaction networks and to predict or understand their dynamics at the molecular interaction level. It is necessary to estimate the unmeasured values of many kinetic parameters for mathematical modeling. Genetic algorithms (GAs) is one of the most efficient method for estimation of the values of kinetic parameters, but large-scale models such as a budding yeast cell cycle model have too many parameters for ordinary GAs to optimize efficiently. To circumvent this problem, we propose a novel evolutionary method based on module decomposition and integration. The large-scale cell cycle model is decomposed into 5 modules. After the optimization of each module, the modules are integrated together and the resultant full model is optimized as multi-objective problems. 2 Method and Results 2.1 Construction of the Cell Cycle Modael Using CADLIVE we drew the budding yeast cell cycle map, as shown in Figure 1. CADLIVE describes not only reactions but also various events such as budding, DNA replication start, spindle formation, and chromatin separation. This map is one of the most sophisticated images for the whole system of the yeast cell cycle [1]. To estimate the kinetic parameter values in this map using genetic algorithms (GAs) efficiently, we decompose it into 5 modules; G1 phase module, S phase (budding) module, S phase (DNA synthesis) module, G2-M phase module, M phase checkpoint module as shown in Fig 1. Figure 1: A budding yeast cell cycle map. (1) G1 phase, (2) S phase (budding), (3) S phase (DNA synthesis), (4) G2-M phase, (5) M phase checkpoint. 2.2 Dynamic Model Regulator-reaction equations are converted into the differential and algebraic equations using the two-phase partition method (TPP). A flow for the conversion from the creation of a network to a dynamic model is performed using CADLIVE [2][3][4]. In simulation of the budding yeast cell cycle, the dynamic model consists of 76 algebraic equations, 40 differential equation, and 116 variables. P066-1 2.3 Optimization We optimized the 5 modules with GAs separately and assumed the top 4 parameter sets in every module as the candidates of optimum solutions, thereby obtaining the assembled model (full model) with 1024 (= 45) parameter sets. We simulated the full model with every candidate parameter set as the initial values for optimization, as shown in Figure 2, to explore the parameter set that best satisfies the experimental data. Since there are various experimental data, the multi-objective GAs (MOGA) are employed. The time course data for the best solution are shown in Figure 3, demonstrating that the cell cycle repeats every 150 minutes [5]. 3 Discussions We estimate the values of kinetic parameters of a large-scale cell cycle model. The optimization method based on module decomposition and integration is effective in optimizing the large-scale biochemical networks. Cell cycle networks have been extensively studied, as it closely relates the elucidation of the molecular mechanisms for cancer cells or general cell growth. The developed mathematical model contributes to advances in the studies of cancer development or systems biology. Figure 2: Image of integration References [1] Katherine C. Chen, Attila Csikasz-Nagy, Bela Gyorffy, John Val, Bela Novak, and John J., Tyson, Kinetic Analysis of a Molecular Model of the Budding Yeast Cell Cycle, Molecular Biology of the Cell, 11:369-391, 2000. Figure 3: The result of simulation [2] Kurata, H., Inoue, K., Maeda, K., Masaki, K., Shimokawa, Y., Quanyu Zhao, Extended CADLIVE: a novel graphical notation for designing a biochemical network map that enables computational pathway analysis, Nucleic Acids Res, 35(20):e134, 2007. [3] Kurata, H., Masaki, K., Sumida, Y., Iwasaki, R., CADLIVE Dynamic Simulator: Direct Link of Biochemical Networks to Dynamic Models, Genome Res., 15: 590-600, 2005. [4] Kurata, H., Matoba, N., Shimizu, N., CADLIVE for constructing a large-scale biochemical network based on a simulation-directed notation and its application to yeast cell cycle, Nucleic Acids Res.31: 4071-4084, 2003. [5] Yamamichi, S., Kurata, H., A large-scale dynamic simulation of the cell cycle network, Proceedings of 7th Asia-Pacific Biochemical Engineering Conference, Jeju Island, Korea, SYS2-05, 2005. P066-2