Computational Method Speeds Mapping of Cell Signaling Networks

For decades, scientists have been studying how external information gets transmitted from outside of cells to the control centers inside them that trigger particular responses. But cell signaling networks are so complex that mapping them has been a slow, arduous process.

Now, a research team from Stanford, MIT and Harvard has developed a new method for charting cellular signaling networks quickly and accurately. Their findings appear in the April 22 issue of Science.

To function properly, cells must be able to detect changes in the environment and respond accordingly--usually by altering their behavior in some way. Cells are equipped with a sophisticated communications network that links surveillance with response, and allows the cells to perform a variety of tasks--such as growing in response to hormones, moving towards a source of nutrients, or responding to foreign invaders in the body. In all of these cases, external signals are converted into cellular messages that get passed sequentially from one molecule to another. Tracing the individual pathways that guide information flow inside cells is a difficult task using standard experimental methods. Multidisciplinary research teams are working on new approaches to make the process faster and easier.

One such group, led by Garry Nolan of Stanford University and Douglas Lauffenburger of MIT has "opened the door for systematic mapping of signaling pathways," according to team member Dana Pe'er. They tested their technique--a combined approach of state-of-the-art experimental and computational methods that "fit like a hand in a glove,"--by mapping a well-studied signaling network in human immune system T-cells. This feat originally took researchers decades to accomplish using traditional approaches.

In the past, scientists have treated signal networks like a puzzle--studying them one pathway at a time and then tying the information together to construct a global picture of what’s going on inside the cell. But the molecules that make up individual pathways in signaling networks rarely operate in isolation--they "talk" to each other and work in sync to process information. What researchers have lacked are the experimental and computing tools to model entire signaling networks at once. This new work should help bridge that gap.

Human T-cells--the system Nolan and Lauffenburger's team used to validate their technique--rely on a complex signaling network to protect the body from pathogens such as viruses. When a T-cell encounters a foreign invader, it relays a danger signal throughout the cell, activating it for a quick response. These signals prompt activated T-cells to multiply rapidly and send out molecular "distress signals" that attract other immune system molecules to battle the infection. Components in the pathway are chemically altered during the signaling.

Such alterations occur only at points of active signaling in the network. The researchers measured the modified signaling molecules to determine where along the pathway T-cell components changed in response to signals. An increase in the amount of the molecules indicated spots where they were most active.

Information about the simultaneous activities of several signaling molecules in thousands of individual cells was analyzed using a powerful computational method called Bayesian network modeling. Bayesian analysis uses statistics to determine associations and dependencies between molecules in a network. With sufficient experimental data, for example, a Bayesian network analysis can not only identify an association between signal molecules A and B, it can detect whether B depends on A for transfer of the signal.

The team's analysis predicted all but three of 18 well-studied T-cell network relationships. Although the work was largely a "proof of principle" report designed to demonstrate the usefulness of their method, their studies also identified a previously unknown T-cell signaling interaction that connects two pathways.

Prediction is the great strength of Bayesian network analysis--particularly in the post-genomic age when researchers are generating vast amounts of data. The statistical approach requires no prior knowledge of signaling pathways to uncover molecular relationships. This is a distinct advantage according to Pe'er, who notes, "Unlike classical biochemical approaches where you need to know all of the intermediate players in a network, this system can detect indirect influences" between molecules.

Pe'er suggests a similar technique could be used to study the differences between normal and abnormal cells. This application would be particularly useful for understanding disease states where signaling malfunction plays a prominent role, including cancer and autoimmunity.

Until now, the large amounts of data the technique requires prevented its application to protein networks. Being able to monitor multiple components in thousands of cells at once was pivotal to the group’s success. Despite technical hurdles, the team believes similar approaches may be used in the future to study systems that involve multiple types of cells and their interactions, such as tissues and organs.

NSF program manager Carter Kimsey thinks this study will open up a wide range of research opportunities. "There is no reason to think that this mathematical tool is for health-related fields only, there are many possible applications in biology," she says.

Dana Pe'er is a recipient of a National Science Foundation postdoctoral research fellowship award in biological informatics. Several organizations supported the work, including the National Institutes of Health, Bristol-Meyer Squibb and the Juvenile Diabetes Foundation.

Source : NSF