Reconstruction of Gene Regulatory Networks by Integrating ChIP-chip, Knock out and Expression Data

Uncovering the underlying regulatory mechanism has become a major research in bioinformatics studies. The availability of various kinds of high-throughput biological data makes the reconstruction of regulatory networks on a genomic scale possible. Since each single data source provides only partial and noisy information of the regulatory relationships, methods combining diverse data sources are expected to get more reliable networks. Here a method was presented to infer the regulatory networks by combining ChIP-chip, TF (transcription factor) knock out and expression data. Since ChIP-chip and TF knock out data provide direct physical binding and functional evidences of relations between TF and target genes, combining these two data is expected to obtain high prediction accuracy. However, the overlap of these two data is low. Based on the assumption that co-regulated genes often have high expression similarity, the method reduced the effect of the low overlap of these two data to some extent. The results show that most inferred regulatory relations are validated by YEASTRACT, high quality ChIP-chip data and literatures, which demonstrate our method is powerful and reliable. Moreover, the comparison between our method and others also shows that it has better performance.