Identification of Gene Regulatory Networks by Integrating Genetic Programming With Particle Filtering

Gene regulatory network can help to analyze and understand the underlying regulatory mechanism and the interaction among genes, and it plays a central role in morphogenesis of complex diseases such as cancer. DNA sequencing technology has efficiently produced a large amount of data for constructing gene regulatory networks. However, measured gene expression data usually contain uncertain noise, and inference of gene regulatory network model under non-Gaussian noise is a challenging issue which needs to be addressed. In this study, a joint algorithm integrating genetic programming and particle filter is presented to infer the ordinary differential equations model of gene regulatory network. The strategy uses genetic programming to identify the terms of ordinary differential equations, and applies particle filtering to estimate the parameters corresponding to each term. We systematically discuss the convergence and complexity of the proposed algorithm, and verify the efficiency and effectiveness of the proposed method compared to the existing approaches. Furthermore, we show the utility of our inference algorithm using a real HeLa dataset. In summary, a novel algorithm is proposed to infer the gene regulatory networks under non-Gaussian noise and the results show that this method can achieve more accurate models compared to the existing inference algorithms based on biological datasets.