``Network science" is a collection of algorithmic, computational and statistical tools that has been developed to study complex systems (systems that are not trivial, regular or completely random). The main premise of network science is that much can be discovered about a system from the underlying topology of the network that describes it. Network science has focused in a wide range of systems, starting from technological networks (the Internet or the Web) to man-made networks (global trade network), to social networks, as well as to ecological networks, gene regulatory networks or metabolic networks in biology. There are several focus areas in network science: large-scale structural methods (e.g., the node degree distribution), microscopic structural methods (e.g., the frequency of certain network motifs), the evolution or rewiring of the network as a function of time, network dynamics (each node is capable of performing a function also based on input from its neighbors), and others.
In this research project, which is joint work with Professor Todd Streelman from the Georgia Tech Biology department, we apply network science methods in developmental biology. Biological development is a temporal and spatial process in which genetic function is marshaled, with extreme precision and control, to make an adult organism. The scientific fascination with development is obvious to any undergraduate (or any parent) when reminded that every human, made of approximately ten trillion specialized cells, begins life as a single-celled embryo. Biologists have studied the details of this process for several organisms and tissues; however, a general theory of biological development has not emerged.
The high-level goal in the proposed research is to understand the development process from a computational perspective, as a distributed algorithm that is executed in a growing and dynamic network of spatially-assigned processors. We consider computational models of development, based on first principles, in which cells correspond to processors, the genetic code of each cell corresponds to the algorithm that each processor executes (identical for all processors), cell actions such as differentiation or death correspond to specific execution stages of the previous algorithm, inter-cell signaling corresponds to communication links between processors, while the developing embryo corresponds to a dynamic and spatial network of processors.
Our models of the development process consist of a dynamic, spatial and growing network. Each network node represents a cell of the developing embryo. The network starts with a single node, the zygote, and it gradually grows to a certain point determined by the developmental process itself. The network is spatial, meaning that each node has a given location, which largely controls the connectivity of that node. Further, the network of cells is dynamic; as new cells are born through cell division, existing cells die through apoptosis. The topology of the network is also subject to rewiring, as cells produce various signals, creating or removing links to other cells dynamically.
Some of the questions we are interested in follow:
Constantine Dovrolis, Spring'09