Reverse Engineering the Cardiogenic Gene Regulatory Network in the Mammalian Heart

<p>Little is known about the underlying regulatory network responsible for the development<br>of the mammalian heart. To address this issue, we utilize our recently developed algorithm<br>to reverse engineer the cardiogenic gene regulatory network using time-series microarray<br>data obtained from the developing mouse heart. The subnetwork ensembles generated by<br>the algorithm consist of topologies capable of explaining the experimental data via model<br>simulation. After pooling these subnetwork topologies together and applying an<br>appropriate cutoff metric to the gene interaction list, a scale-free, hierarchical network<br>emerges. The network is validated with known gene interactions and used to identify new<br>regulatory interactions and network hubs critical to the developing mammalian heart. The<br>candidate gene interactions identified by the algorithm is prioritized using semantic<br>similarity and gene profile uniqueness metrics to produce a list of testable interaction<br>pairs. Among the top 25 gene pairs identified, significant fraction have already been<br>validated. Furthermore, the network was expanded using the same semantic similarity<br>and gene profile metrics to include all genes in the mouse genome to form the most likely<br>cardiogenic gene regulatory network predicted by the algorithm. The method outlined<br>herein provides an informative approach to network inference and leads to clear testable<br>hypotheses related to gene regulation. Massively parallel architecture of GPUs have been<br>tested to study genome-scale problems.