Inferring gene regulatory networks using a hybrid GA–PSO approach with numerical constraints and network decomposition

Gene regulatory networks (GRNs) are essential for cellular metabolism during the development of living organisms. Reconstructing gene networks from expression profiling data can help biologists generate and test hypotheses to investigate the complex phenomena of nature systems. However, building regulatory models is a tedious task, especially when the number of genes and the complexity of regulation increase. To automate the procedure of network reconstruction, we establish a methodology to infer the computational network model and to deal with the problem of scalability from two directions. The first is to develop an enhanced GA–PSO hybrid method to search promising solutions, and the second is to develop a network decomposition procedure to reduce the task complexity. Meanwhile, our work includes a quantitative method to consider prior knowledge in the inference process to ensure validity of the obtained results. Experiments have been conducted to evaluate the proposed approach. The results indicate that it can be used to infer GRNs successfully and can achieve better performance.