Multiscale Modelling of Gene-Regulation Networks

The Computational Biology application in MAPPER is concerned with developing and deploying computational biology tools on the MAPPER e-infrastructure focusing gene regulation. Our vision is to enable modelling and simulation of large gene-regulation networks that exhibit an inherent sub-division particularly in the time dimension. Such gene networks typically consist of several dozens of genes. They are difficult to handle with conventional modelling and simulation approaches due to the conceptual and computational complexities involved. Currently, we explore various aspects of reverse-engineering gene-regulation networks (GRNs) from gene expression data. Of particular interest are GRNs exceeding 20 genes, reverse-engineering without known GRN topology and from multiple condition experiments, methods for resampling the same data sets for model construction, assessment of structure recovery (as opposed to behaviour prediction), and distribution and scalability of the underlying computations. Our current approach views these tasks as being "multiscale" in the sense that the optimization procedure underlying the reverse-engineering process is partitioned into multiple optimization "islands" within a particle swarm optimization framework. We have developed a Java-based GRN modelling and simulation software library called "Multiscale Gene Regulation Modeling Tool" (MultiGrain) and are currently undertaking various computational studies into the GRN problems mentioned above.