Different superfamilies are formed by microbial transcription networks (top) and developmental and neuron organization networks (bottom).

Alon/AAAS

Networks from very different systems fall into just a few superfamilies, say Ron Milo, Uri Alon, and colleagues (Weizmann Institute of Science, Rehovot, Israel). The superfamilies are based on the relative frequencies of certain local network motifs or recurring circuit elements that have defined information processing tasks.

Alon says he wants to “break the network down into elementary circuits and building blocks. We're very much inspired by engineering. If you want to understand the device…you break the problem up.”

He has previously found that networks have higher than expected frequencies of certain circuit elements; such elements have functions such as detecting persistence or imposing temporal order. The team now quantifies the relative occurrence of such elements in a wide variety of networks and finds that just a few superfamilies emerge.

The first includes transcription networks from microorganisms. These sensory networks must respond rapidly relative to the buildup time for each of their components, and thus feature a shallow organization. By contrast, developmental transcription networks in flies and neuron organization in worms have an output time (hours or a second) that greatly exceeds the time taken for each step (minutes or 100 ms). Therefore, they have less bias against longer cascades and feature more feedback loops; these non–rate-limited networks form another superfamily.

The families are not restricted to biology—internet links and social networks form another superfamily, and diverse languages a fourth. All this clustering offers exciting opportunities for understanding, for example, fly development by taking lessons learned from worm neuronal architecture.

The current study covers only circuit elements with three or four components. Alon is currently pushing this number higher. “How we can ever cope with the complexity of these systems,” he says, “is to work at the level of individual circuits rather than individual arrows.” ▪

Reference:

Milo, R., et al.
2004
.
Science.
303
:
1538
–1542.