1 Temporal Dynamics of Regulatory Networks in Drosophila melanogaster Embryogenesis

Dynamic modeling of regulatory networks that control gene expression requires temporal information on the activation/repression latencies of regulator and target pairs, which have been experimentally inaccessible at the genome scale. We developed a new discretization method using multi-step functions, to systematically infer latency information for individual edges of a large-scale regulatory network from whole-genome time-course expression profiles. Our method has wider applicability and shows increased accuracy relative to previous approaches such as Pearson or Spearman correlation. It also exhibits good predictive power of expression co-localization for regulator/target pairs benchmarked against the ImaGO annotation for Drosophila melanogaster. Application of this method to D. melanogaster led to several new insights on its network dynamics. First, the measured delays are significantly longer than expected by chance, and are specific to D. melanogaster. They are not found in yeast, suggesting that they are likely relevant to animal genomes and developmental processes. Second, we found that regulator binding site multiplicity on the target promoter region is related to an increased latency, which is consisten