Poster: Comprehensive pharmacogenomic pathway screening by data assimilation

Construction and simulation of biological pathways are crucial steps in understanding complex networks of biological elements in cells. To construct simulatable models, structures of networks and chemical reactions are collected from existing literature and the values of parameters in the model are set based on the results of biological experiments or estimated based on observed data by some computational method. However, it is possible that there are some missing relationships or elements in the literature-based networks. In this paper, a method that can create a set of extended simulatable models from prototype literature-based models is focused on. Biological simulation models were formulated under a framework of nonlinear state space model in order to use observed data for parameter estimation. There are two key points in the proposed strategy: One is that various structures of candidate simulation models are systematically generated from the prototypes. The other is that, for each created model, the values of parameters are automatically estimated by data assimilation technique ; the values of parameters will be determined by maximizing the prediction capability of the model. For the comparison of multiple simulation models, Bayesian information criterion (BIC) was employed.