Integrative approaches to study TF-DNA interactions

Transcriptional regulation is a crucial process for cell differentiation and development. This process is tightly coordinated by the binding of key proteins, called transcription factors (TFs), to the DNA in a sequence-specific manner. The binding of TFs at their TF binding sites (TFBSs) ensures that the right genes are expressed in the right tissue and at the right levels. Alterations occurring at TFBSs, due to mutations or other genomic alterations, can disrupt the transcriptional regulatory mechanism, affecting entire gene regulatory networks, which in turn can lead to diseases such as cancer. The aim of the thesis was to develop and improve computational tools and methodologies for genome-wide identification of TFBSs, capitalizing on the large amounts of publicly available ChIP-seq data (the most popular experimental assay capturing TF-DNA interactions in vivo), and use this information to infer gene regulatory networks. Employing a non-parametric, entropy-based computational approach, we improved our capacity to identify bona fide TFBSs. As a result, a map of direct TF-DNA interactions in the human genome, hosting ~8 million binding sites for more than 230 TFs was generated and made publicly available. This resource, together with other relevant genomic information, served as a base for improving the prediction of TF target genes, called regulons, furthering the identification of gene regulatory networks. Altogether, these resources were put to use in a practical setting, identifying candidate TFs involved in breast cancer development, discriminating between estrogen receptor positive and negative breast cancer subtypes. In brief, this thesis presents the development of new computational methods and resources that are derived from in depth analyses of experimentally-generated data to study gene expression regulation and how it can be disrupted in diseases such as cancer