A cubature Kalman filter approach for inferring gene regulatory networks using time series data

Abstract

A novel technique for the inference of gene regulatory networks is proposed which utilizes cubature Kalman filter (CKF). The gene network is modeled using the state-space approach. A non-linear model for the evolution of gene expression is considered and the microarray data is assumed to follow a linear Gaussian model. CKF is used to estimate the hidden states as well as the unknown static parameters of the model. These parameters provide an insight into the regulatory relations among the genes. The proposed algorithm delievers superior performance than the linearization based extended Kalman filter (EKF) for synthetic as well as real world biological data.