Inferring a Transcriptional Regulatory Network from Gene Expression Data Using Nonlinear Manifold Embedding

Transcriptional networks consist of multiple regulatory layers corresponding to the activity of global regulators, specialized repressors and activators of transcription as well as proteins and enzymes shaping the DNA template. Such intrinsic multi-dimensionality makes uncovering connectivity patterns difficult and unreliable and it calls for adoption of methodologies commensurate with the underlying organization of the data source. Here we present a new computational method that predicts interactions between transcription factors and target genes using a compendium of microarray gene expression data and the knowledge of known interactions between genes and transcription factors. The proposed method called Kernel Embedding of REgulatory Networks (KEREN) is based on the concept of gene-regulon association and it captures hidden geometric patterns of the network via manifold embedding. We applied KEREN to reconstruct gene regulatory interactions in the model bacteria E.coli on a genome-wide scale. Our method not only yields accurate prediction of verifiable interactions, which outperforms on certain metrics comparable methodologies, but also demonstrates the utility of a geometric approach to the analysis of high-dimensional biological data. We also describe the general application of kernel embedding techniques to some other function and network discovery algorithms