Reconstructing Gene Regulatory Network Using Heterogeneous Biological Data

Gene regulatory network is a model of a network that describes the relationships among genes in a given condition. However, constructing gene regulatory network is a complicated task as high-throughput technologies generate large-scale of data compared to number of sample. In addition, the data involves a substantial amount of noise and false positive results that hinder the downstream analysis performance. To address these problems Bayesian network model has attracted the most attention. However, the key challenge in using Bayesian network to model GRN is related to its learning structure. Bayesian network structure learning is NP-hard and computationally complex. Therefore, this research aims to address the issue related to Bayesian network structure learning by proposing a low-order conditional independence method. In addition we revised the gene regulatory relationships by integrating biological heterogeneous dataset to extract transcription factors for regulator and target genes. The empirical results indicate that proposed method works better with biological knowledge processing with a precision of 83.3% in comparison to a network that rely on microarray only, which achieved correctness of 80.85%.