Quantifying transcriptional regulatory networks by integrating sequence features and microarray data.

Genome-wide transcriptional regulatory networks (TRNs) specify the interactions between transcription factors (TFs) and their target genes. Many methods have been proposed to reconstruct regulatory networks from gene expression datasets and/or genome sequences, but most of them can only infer qualitative regulation relationships. Thus, developing a quantitative model that can estimate the kinetic parameters of transcriptional regulatory functions is an urgent and important task. In this paper I propose REMBE, a regulatory model based on binding energy, to quantify transcriptional regulatory networks. My model combines multiple kinetic quantities, including binding strength, TF-DNA's binding energy, transcription productivity with respect to each binding state, and hidden TFs' concentration, into a general learning model. Experimental results show that my model can effectively learn these kinetic parameters and TFs' concentration from genome sequences and gene expression data. Moreover, these learned parameters and TFs' concentration provide more informative biological senses than merely qualitative regulatory relationships can do.