Gene regulatory networks of transcription factors (TF) and their target genes, implemented through TF binding sites, are key features of biology. Their characterization poses a challenge to regulatory bioinformatics, namely to develop and evaluate methods that can generate high accuracy lists of candidate target genes for each transcription factor in the network. In a handful of cases, using sequence conservation or binding site clustering, has genome-scale scanning for regulatory interactions been successful in higher eukaryotes. However, not all regulatory regions are conserved at sequence level, not all binding sites occur in tight clusters, only few TFs have a high quality PWM, and for only a few TFs the cooperating factors and their PWMs are known. Therefore, it is currently unclear which paradigms can be applied for which TFs to yield successful target predictions, and how multiple genomes can be optimally exploited to this end.

In the first part I will present a benchmark study on 34 TFs in Drosophila to assess existing TF binding site and cis-regulatory module discovery and scoring methods, taking advantage of the availability of multiple Drosophila genomes. From this, a novel methodology is derived, called cisTarget, to assess and utilize the optimal TG discovery strategy for each individual TF. In the second part I will present a method for target gene prioritization that integrates multiple heterogeneous data sources on functional gene annotation, pathways, literature, protein sequences, protein-protein interactions, genetic interactions, and gene expression profiles.