Prediction of cardiac transcription networks based on molecular data and complex clinical phenotypes

We present an integrative approach combining sophisticated techniques to construct cardiac gene regulatory networks based on correlated gene expression and optimized prediction of transcription factor binding sites. We analyze transcription levels of a comprehensive set of 42 genes in biopsies derived from hearts of a cohort of 190 patients as well as healthy individuals. To precisely describe the variety of heart malformations observed in the patients, we delineate a detailed phenotype ontology that allows description of observed clinical characteristics as well as the definition of informative meta-phenotypes. Based on the expression data obtained by real-time PCR we identify specific disease associated transcription profiles by applying linear models. Furthermore, genes that show highly correlated expression patterns are depicted. By predicting binding sites on promoter settings optimized using a cardiac specific chromatin immunoprecipitation data set, we reveal regulatory dependencies. Several of the found interactions have been previously described in literature, demonstrating that the approach is a versatile tool to predict regulatory networks.