Modeling gene regulatory networks based on nonlinear state-space model

In the gene regulatory networks(GRNs), gene expression is usually regulated by some regulatory factors (or transcription factors), but the regulatory factors' activity is difficult to measure, and the regulatory effect among genes is typically nonlinear. This paper uses nonlinear Gaussian state-space model to construct gene regulatory networks. In the model, genes are considered as observation variables and regulatory factors are considered as internal state variables, it is more consistent with the actual biological systems. To identify the system, the unscented Kalman filter(UKF) algorithm is applied to estimate the states and parameters, Bayes information criterion (BIC) is applied to determine the dimension of the state variables. We model the regulatory networks of yeast genes' expression data with the method, the results show that the nonlinear state-space model can improve the accuracy of constructing regulation networks, and UKF algorithm can effectively estimate the parameters and states of nonlinear state-space model.