Inference of Gene Regulatory Networks with Local Lasso Path Consistency Algorithm Based on Conditional Mutual Information

In recent years, gene regulatory networks (GRNs) are becoming one of the most challenging problems in systems biology. Capturing the interactions among genes from analyzing gene expression microarray data could help us understand the regulatory mechanisms in cells or genomes. Various methods based on information theory have been applied to infer the GRNs. However, due to the nonlinear relationship between genes, the sparsity in the network structure and the external noise in the original data, these methods more or less bring out the redundant regulatory links in the process of network reasoning. Especially, with the increase of network scale, there is still plenty of room for us to improve their performance and efficiency. In this paper, for the path consistency algorithm based on conditional mutual information (PCACMI), local Lasso path consistency algorithm based on conditional mutual information (Loc-Lasso-PCACMI) is proposed, which integrates Lasso algorithm with PCACMI from the view of local inference. Based on the benchmark data sets from the DREAM challenge and the simulated data sets from the SynTReN network generator, all of the results demonstrate our proposed method is superior to other classic methods of information theory.