Skip to Main Content

Computer Science|College of Engineering and Applied Science

UW Professor Helps Discover 'Holy Grail' Theory of Evolving Modular Networks


January 30, 2013 - One of the University of Wyoming's newest professors has tackled a decade-old question regarding the evolutionary origins of modularity and discovered a new theory for why biological entities are organized into modules.

While a visiting professor at Cornell University, Jeff Clune and fellow researchers discovered this: Modularity in biological networks, such as networks of neurons and genes, does not evolve because it speeds up adaptation, as the current leading theory professes. Rather, modularity in such organisms evolves because it saves on "wiring costs" due to modular networks' use of fewer and shorter connections.

"Neurons mostly connect only to nearby neurons," says Clune, an assistant professor of computer science who became a faculty member at UW this month. "Evolution doesn't want to build tissue it doesn't need, so it saves on the cost of building and maintaining neural connections by using a modular design."

This new theory is detailed in a paper titled "The Evolutionary Origins of Modularity," which was published in the Proceedings of the Royal Society Jan. 30 (today). In addition to Clune, the paper's authors include Hod Lipson, an associate professor in Cornell University's departments of mechanical and aerospace engineering and computer science; and Jean-Baptiste Mouret, a robotics and computer science professor at Universite' Pierre et Marie Curie in Paris, France.

Many biological entities -- from human brains to gene regulation to protein interactions -- are organized into modules, which are essentially dense clusters of interconnected parts within a larger network. These modular designs are similar to how engineers build cars out of various components, such as spark plugs and fuel injectors, or how children create structures from Lincoln Logs or Legos.

Using powerful computers, the research group simulated 25,000 generations of evolution. They were able to test their theory by evolving networks with and without a cost for network connections.This computer model illustrates a network of neural connections in a human brain.

"Once you add a cost for network connections, modules immediately appear," Clune says. "Without a cost, modules never form. The effect is quite dramatic."

Clune likened the findings to a road network. Within a city, many of the roads are connected to each other, but few of the local roads are connected to the roads in other cities. Instead, you have just a few connections between cities (e.g. highways). Biological networks are organized similarly, and for a similar reason: it is expensive to build connections (roads), especially long ones.

Clune and his team used computational simulations of evolution for their study for two reasons. It is much faster than natural evolution, allowing evolutionary experiments with thousands of generations to occur in a few days, and it provides more experimental control. It would have been impossible, for example, to conduct this experiment in a naturally evolving species because there is no way in nature to eliminate the cost for connections.

In addition to helping biologists understand why organisms are built in a modular fashion, the results may well have significant implications for evolutionary computation, which is a field that harnesses evolution for engineering purposes, such as evolving artificially intelligent robots.

"Trying to figure out how to evolve modularity has been one of the 'holy grails' of this field," says Clune, whose specialty is artificial intelligence. "We can use this discovery to create more intelligent robots that can find people stranded in an avalanche, pick up trash in national parks or disarm landmines."

The research was supported by the National Science Foundation and the French National Research Agency. To read the paper, go to http://rspb.royalsocietypublishing.org/content/280/1755/20122863.full.

News Coverage
National Geographic: The Parts of Life,
MIT Technology Review: Computer Scientists Reproduce the Evolution of Evolvability,
Cerbral Hack: Scientists crack the code as to how modular design can evolve,
Science Recorder: Researchers simulate 25,000 generations of evolution, boost artificial intelligence,
Science A GoGo: Simulation reveals evolutionary origins of modularity,
Science World Report: Researchers Find Secret Behind Biological Modularity; Artificial Intelligence Gets a Boost,
Science Daily: Engineers Solve a Biological Mystery and Boost Artificial Intelligence,
Newswise: Cornell Engineers Solve a Biological Mystery and Boost Artificial Intelligence,
E! Science News: Cornell engineers solve a biological mystery and boost artificial intelligence,
Phys Org: Researchers solve biological mystery and boost artificial intelligence,
Chronicle Online: Scientists find 'holy grail' of evolving modular networks

Photo:
Jeff Clune, a UW assistant professor of computer science, was part of a research team that discovered that modularity in humans and other organisms evolves because it saves on "wiring costs" -- the costs associated with building and maintaining network connections encourages the evolution of modular designs, which have fewer and shorter connections.

Footer Navigation

University of Wyoming
 
1000 E. University Ave. Laramie, WY 82071 // UW Operators (307) 766-1121 // Contact Us