Published online 6 August 2008 | Nature | doi:10.1038/news.2008.1010

News

Google tool identifies linchpin species

Search system predicts what prey are needed to keep an ecosystem working.

Google’s search algorithm can be used to determine which prey are most important for an ecosystem to thrive.

That’s the claim of a researcher who studies food webs, the complex networks that describe who eats whom in an ecosystem. The more complex the web, the harder it is to determine what would happen if various prey were removed from the ecosystem.

gazelleThe PageRank algorithm could help to rate the gazelle's importance in its food web.Punchstock

The huge number of possible sequences means that “the space of possible solutions is practically infinite”, says Stefano Allesina of the National Center for Ecological Analysis and Synthesis at the University of California, Santa Barbara.

The gold-standard method for dealing with what happens when a prey species is removed from the system is to use a genetic algorithm in which randomly generated possible solutions are assigned a greater or lesser degree of fitness. Fitter solutions are then combined to create a next-generation algorithm that benefits from their superior elements. The idea is to ultimately ‘evolve’ an optimal solution.

But running such an algorithm on a complex network is time-consuming and expensive, and Allesina wanted something simpler. So he turned to Google’s formula.

PageRank power

The magic behind Google’s search prowess is an algorithm called PageRank. The formula assigns pages with a rank according to how many other pages link to it, and what rank those pages have.

To apply the idea to food webs, Allesina coined a central rule: ”a species is important if it points towards important species.” In other words, you are important prey if important predators eat you.

Allesina also made a crucial tweak to his food web so that his version of PageRank could work on it. He made the network irreducible, ensuring that there were no dead-ends or islands.

At first glance, it seems as though food webs just don’t work this way. Gazelle eat grass; lions eat gazelle; yet few animals eat lions, making them a potential dead-end.

Allesina found a solution he delicately calls ”detritus”, which involves installing a 'root node' that represents species that feed on the droppings and carcasses of species higher up the food chain. That ensures that the lion is connected as 'prey' to the root node, which is in turn connected to the grass.

Top ranking

After a bit of fine-tuning, Allesina ran his algorithm on 18 data sets that had already been number-crunched by the conventional algorithm, and found that his results were “statistically indistinguishable” from those produced by the gold standard.

The result is a ranked list of which species to remove — and in which order — to get maximum ecosystem meltdown. It sounds grim, but it could be very useful to guide conservation efforts toward the key species that need protecting to maintain an ecosystem.

Allesina now hopes to include top-down causes of extinctions such as habitat loss, competition from invasive species and low birth rates, in his analysis. He presented his results at the Ecological Society of America meeting in Milwaukee, Wisconsin, on 4 August.

“This looks like quite a sensible approach to making sense of food web structure,” says Carl Bergstrom, a biologist at the University of Washington in Seattle, who uses Google’s algorithm to rank journals by calculating their impact factor per dollar at eigenfactor.org. “Eigenvector centrality methods, such as PageRank, can be extremely effective at extracting global information from network structure. It’s very good to see people exploring some of the more sophisticated network measures to address problems in community ecology.” 

Commenting is now closed.