Ever since Seattle's Institute for Systems Biology was founded in 2000, co-director Alan Aderem and his colleagues have been gearing up for the type of experiment published on page 173 of this issue. They are working to elucidate the complex networks that regulate the body's immune system. And they use both computational and experimental approaches, with each of the techniques informing and validating the others.

Aderem is particularly interested in finding out how the immune response is regulated by macrophages — a group of white blood cells that fight foreign particles such as bacteria. These cells produce a slew of substances, called cytokines, which target invaders and call other immune cells into action. Once the danger passes, the macrophages help to call off the immune response that, left unchecked, can damage the body's own tissues. But how do macrophages orchestrate this regulation?

To get to the answer, Aderem and his colleagues opted for a ‘systems biology’ approach. Starting with a clean slate, they investigated the entire cell machinery rather than just focusing on specific aspects of it. The overall strategy was developed during discussions with other faculty members. “We would ask: how can we approach this kind of problem? What tools to we need?” says Aderem.

The first step was to use microarray chips to identify various sets of genes turned on at different times after mouse macrophages were exposed to a bacterial toxin. The team suspected that any transcription factors in the first set of genes would turn on genes in the second set, and transcription factors in that set would turn on genes in the third, and so forth.

Armed with this assumption, they used two computational tools, called MotifMogul and Cytoscape, to identify the transcription factors within the first set of expressed genes that worked in concert to turn on specific genes in the second set. In this way, they began to build molecular networks, showing how different molecules interact to turn the immune response on and off.

To test whether their strategy was working, they focused on specific molecules within the networks and tested their functions in the immune system using a range of experimental techniques, including knockout mice. Biologists are sometimes sceptical of computational approaches. But the key, says Aderem, is that “computational data are validated by experiments”. And experiments, in turn, improve computational tools. “MotifMogul is a program that learns, so it will work better next time,” he says.

Scientists at the Seattle institute developed the necessary computational and experimental tools over the past six months and validated them using simpler organisms such as yeast. The assumption they were banking on was that transcription activates subsequent waves of transcription factors. But in some situations, protein phosphorylation or other mechanisms regulate transcription. For such systems, other tools would be needed to build networks. “In this case our strategy worked,” says Aderem, “it hasn't in others.”

“Our paper shows our approach is extremely powerful,” adds Aderem. “If we understand clearly how immune responses unfold, we can perturb them selectively.” And that could eventually lead to cures for autoimmune diseases and more effective vaccines.