Data-Driven Approximation of Gene Regulatory Networks Using Extended Dynamic Mode Decomposition

This paper addresses the data-driven modeling problem of a class of gene regulatory networks. Utilizing the Koopman operator theory, the study applies the extended dynamic mode decomposition (EDMD) method to linearize a high-dimensional nonlinear biochemical network model, and extends the state model establishment without input to the model establishment with input. Finally, the control signal obtained by the model predictive control (MPC) method is used to converge the state of the original nonlinear model. The results show that the data-driven method applied in this study can effectively establish the linearized model of the nonlinear system and make the subsequent linear control feasible.