Under the Microscope

Disease 'networks' mimic web

Studying the spread of virulent computer viruses may prove useful in understanding the spread of disease and the ability of ecosystems to handle disturbances, researchers say.

"In terms of computer networks, one of the clear points that comes out of the analysis is that if you have this scale-free network, most transmission can be traced to the most highly connected nodes. So this is a clear implication for how to prevent the transfer of viruses: You concentrate on the most highly connected nodes," says Alun Lloyd, a researcher at the Institute for Advanced Study. Lloyd and Oxford Universitys Robert May reported their findings in the journal Science on May 18.

Computer and biological networks have similar structures that affect how disturbances such as electronic viruses propagate through them. Computer networks are "scale-free" networks, meaning most nodes of the network have relatively few connections to other nodes, while a small number have many connections.

For instance, a university, Internet service provider or large company like Microsoft will have thousands or millions of connections to other points in the network, while a home computer may have only one. So, a virus that hits an individuals PC is likely to propagate more slowly than one that invades Microsoft, since there are fewer links to exploit.

The computer case mimics what happens in the spread of diseases in the real world. With sexually transmitted diseases like AIDS, "a few individuals — such as prostitutes — have very high numbers of partners," the researchers wrote.

On the ecological front, so many processes are involved that the conclusions are not so clear. The model might be used to develop plans for protecting endangered species. "A food web may be one of those networks, so there are interactions in species where the nodes are the species and the links are that one species eats this one and competes with another one," Lloyd says. "The stability of the ecosystem might depend on these links.

"With some species," Lloyd says, "you could remove it quite easily and it might not have much of an effect. But a species with a lot of links might have a very large effect."

A few years ago, IBM scientist Jeffrey Kephart anticipated that the study of this interconnectedness — called topology in mathematics — might yield important theoretical conclusions about population biology and epidemiology. "For example, in this heyday of HIV, we are admonished daily by educators about the dangers of promiscuous activity, yet until recently there were no quantitative theoretical studies of how the spread of disease depends upon the detailed network of contacts between individuals," Kephart wrote.

"Digital organisms" may be preferable subjects in the study of disease, he wrote, because they can be more easily controlled experimentally.

Lloyd and Mays findings differ somewhat from the conclusions drawn in related research and reported in Physical Review Letters by Romualdo Pastor-Satorras and Alessandro Vespignani. Even at very low levels of infection, a computer virus will spread widely, they concluded in their paper

But Lloyd says their work used a model in which the infected node could be reinfected and continue to spread the virus — the Typhoid Mary of computer viruses. In humans, with most viruses, once a person is infected, he or she gains some immunity. In computers, the most highly connected nodes are usually those that are most sophisticated in dealing with virus infections, he says.