biological guild project

Discovering Biological Guilds Through Topological Abstraction

In the last several years, biomedical informatics has undergone dramatic change, as new information has been infused from the human genome and other high throughput biological datasets. Much of this data is now starting to involve relationships- between genes through their regulation, between proteins through their interactions. While in the past, genes were studied one at a time, we can now look at multiple genes and their resulting proteins in parallel. This has lead to fantastic networks- such as gene regulatory networks- where nodes represent genes [point] and lines between the nodes (or edges) [point] represent gene regulation. In gene regulation the state of one gene affects the state of another.

Such networks are often referred to as “hair balls”- they are too large to interpret visually and too complicated to be easily analyzed. Thus, investigators have sought to look inside such networks by looking at small subsets of the global network using local properties- that is looking at a given node and its neighbors. Hubs have been surprisingly powerful in giving us biological insights into networks. It has been shown that hubs predict biological behavior. Taking out an important node such as a hub can result in cell death. Yet, looking at local pathways and hubs only gives us part of the picture. To capture emerging properties of the network as a whole the approach taken is to go beyond local properties and pathway analysis- which effectively yield anecdotes about the network and is to create an abstraction of the network

that looks at it as a whole and captures its core connectivity- by consolidating relationships to reorganize the network. The focus is on the problem of analyzing complex biomedical networks and applied approach to this - an abstraction process to effectively transform indirect connectivity patterns into direct links. These local connections can then be easily analyzed and visualized. The results, including the discovery of a new type of biologically relevant topological class, which can be called a guild- a group of highly interconnected nodes, or clique, with similar function and its correspondence to a specific biological function. However significance of this approach is results and findings that could not be derived from solely looking at direct links in the original network.