Diana Steele's Home Page | Science Writing | Previous Story | Next Story

Scientists search for secrets 
of robust systems
Multiple features help life, technologies survive disturbances

Engineers and designers have long struggled to find ways to make their products -- like planes -- more robust. 

September 18, 2000 

By Diana Steele / Special Contributor to The Dallas Morning News

When the economy goes sour, companies that make more than just widgets are more likely to survive. When a meteor hits the Earth, families of creatures with more diversity among their members are more likely to have survivors after the dust clears.

Planes still fly even if one rivet pops off a wing. Computers still work (for the most part) even if software has a few bugs in it. And power grids, in the United States at least, are pretty reliable at providing power 24 hours a day. 

These systems all share in common a quality called "robustness" ó an ability to keep functioning even when a part goes awry or something in their environment changes. Airplane designers, computer makers and electric companies have long struggled to find ways to make their products more robust. 

Now scientists are getting into the game and trying to pin down some general principles. Robustness is emerging as a discipline of its own, but itís still an inexact science. Defining precisely what it means is the first hurdle.

Courtesy Mike Clark, Cambridge University

The immune system's robustness depends on the remarkable diversity of this protein -- immunoglobin. The key is the mutability of the molecule's two "arms."

"Everybody has an intuitive feeling as to what it means," says Erica Jen, a mathematical biologist at the Santa Fe Institute. But "nobody really has even a working definition of what it means precisely. Everybody has a gut-level feeling that itís key to survival and itís important for performance. And itís important for adaptability and evolvability. But people use the word differently in different contexts."

Robustness is on the verge of emerging from "mystical notions," says John Doyle, a  professor of control and dynamical systems at the California Institute of Technology. Robustness now seems about as mysterious as energy did in the early days of steam engines or electricity, he says. But thereís nothing really magical about it. 

The need to understand robustness is the inevitable consequence of technological progression from building parts to building systems. "We know how the parts work, to a great extent," Dr. Doyle says. "And we can build practically anything you could imagine." But understanding how to put all the parts together and make them function in a robust way ó "thatís the dominant issue facing complex systems."

Scientists at the Santa Fe Institute, funded by a new three-year, nearly $1 million grant from the Packard Foundation, hope to bring together scientists who are studying robustness in different disciplines. 

 "The approach weíre going to take is to start from biological systems," says Dr. Jen, "in which we actually have a hope of pinning down what it might mean for a cellular process to be robust." While robustness is only relatively recently emerging as an issue in engineered systems, biological systems have had eons of evolution to figure it out.

Cells, for example, like neurons in the brain, keep transmitting information reliably even when they have to contend with a noisy or fluctuating environment. This is one type of robustness in biological systems that translates directly to engineered systems, like computer networks, for example.

Another area that Santa Fe scientists will focus on is robustness against mutation. They hope to examine how robustness against mutation on the one hand lessens an organismís vulnerability to its environment. But on the other hand, susceptibility to mutation increases an organismís ability to evolve and adapt to new challenges ó another form of robustness.

One system that takes advantage of high mutation rates to achieve robustness is the immune system. 

Antibodies called immunoglobulin act as the immune systemís foot soldiers in the fight against infection. They stick to foreign molecules that make their way into the body and flag them for destruction. But in order to adapt to the wide variety of incoming germs, these antibodies have to be extremely diverse.

Santa Feís Thomas Kepler, a mathematical biologist and vice president of the institute, says the rate of mutation among the cells that produce these antibodies is extremely high: about one mutation for every cell division. So each time a cell divides and reproduces, it has the ability to produce a slightly different antibody.
"The problem with this is that most mutations either donít help at all, or hurt the molecule, prevent it from doing its job," says Dr. Kepler. 

He says the immune system gets around this problem in two interesting ways. First, before the cells start mutating, they reproduce normally in very large numbers. 

"If you separate growth from mutation, you can achieve very significant growth of the cell population, and then switch mutation on," he says. "And when you do that, you have a very large population of cells to mutate, and they are able to explore the space  [mutation possibilities] much more efficiently and without fear of losing everything." 

Secondly, some portions of the genes that code for these antibodies are much more susceptible to mutation than others. It turns out that portions of the code that are important for the overall structural framework of the molecule mutate much more slowly than the part that normally sticks on to foreign bodies. "Thatís where the changes might potentially do good," says Dr. Kepler. More mutations in the portion of the molecule that latches onto invaders mean more possibilities for fighting disease.

The immune system illustrates some principles that seem to be common to robust systems. It tailors itself to changing enemies ó itís adaptable. If part of the system fails, the whole thing doesnít go down ó itís modular. It has a wide range of defensive weapons ó itís diverse. And there are backup plans if the first ones go awry ó itís redundant. 

Scientists are exploring how these concepts apply in other systems, both natural and designed.

For example, in learning how to preserve ecosystems, understanding robustness is crucial. Simon Levin, a Princeton University ecologist and author of the book Fragile Dominion, worries that the homogenization of agriculture is producing ecosystems that are less robust. Agriculture in which one species, like corn, for example, is grown over vast areas may be especially vulnerable to changing environmental conditions or to the outbreak of a new pest.

He says diversity is a crucial part of a robust ecosystem. But the best way to maintain that diversity is hotly debated. A greater degree of interconnectedness between ecosystems, for example, may prevent species from dying out due to isolation. (Isolated populations tend to become genetically inbred and more susceptible to life-threatening mutations.) On the other hand, this greater connectedness   ó less modularity ó may allow diseases to spread more rapidly.

"Those are the kind of research questions people are wrestling with now, in the design of nature reserves," he says. "How do we design the structure of a system so as best to maintain the populations of interest? There is no single answer."

Heís also interested in learning whether changes can portend whether a system is healthy or teetering toward collapse. "Can we read into changes in the structure of a system some hints that the system is tending toward more vulnerable states?" he asks ó be it a disease outbreak or a stock-market collapse. "And if so, can we manipulate systems to get them away from the abysses, from the edges?" 

Caltechís Dr. Doyle maintains, "With respect to robustness, we are a primitive culture." He says in the early days of electricity, there was a lot of mysticism about energy: Because people didnít understand how electricity worked, notions about its generation and transmission were vague or incorrect. But todayís technological culture takes all that for granted. Now, Dr. Doyle says, there are a lot of mystical notions about robustness, which is at best "still a subject in it infancy."

The key thing to understand about robustness is that "itís a trade-off," says Dr. Doyle. "Well-designed systems are all robust and fragile" at the same time. "Strictly speaking, itís not right to say one thing is more robust than another. You can think about how efficiently something achieves robustness."

For example, in the days before airbags, high-speed, head-on collisions were more likely to kill a carís occupants. Now, airbags mitigate that danger. But at the same time, airbags make people more vulnerable to death or injury from an airbag deploying spontaneously in a parking lot, for example. "So you get killed in a circumstance that without an airbag would be completely safe," he says. "Youíve shifted the vulnerability, and if you do it in a systematic way, itís a net win."

He said it would be possible to design an airbag to deploy differently during low-speed vs. high-speed collisions, or to be sensitive to the weight of a person in the front seat ó a child, for example. But introducing new vulnerabilities during the process would be unavoidable. 

For all the robustness that high mutation rates lend the immune system, tumors in the cells that produce immunoglobulin are more common than many other types of cancer. "This seems to be a penalty we suffer in order to allow these
cells to 'mutate' at a high rate," says Michael Clark, an immunologist at Cambridge University in England. "Thus whilst the mutations allow us to
survive infection better, if these mutations are not controlled properly
we run the risk of other self-inflicted diseases."

Dr. Doyle says these types of trade-offs are mathematically inevitable. "All those complexities introduce new trade-offs," he says. "If you understand those trade-offs, you can design them in such a way that it all works out. If you donít, then you could be in real trouble, and the worst thing is, if you really forget something Ö if thereís some substantial vulnerability that you create that youíre not aware of."

Humans, for example, are much more robust to temperature extremes like a Texas summer and a Chicago winter. But simpler organisms, like bacteria, tend to be adapted to very narrow niches. Some have to live at human body temperature, and will die if exposed to boiling water. Others thrive at the extremely high temperatures and pressures found at deep-sea vents, only to die when brought to the ocean surface.

But in the event of a huge meteor striking the Earth, as is thought to have happened 65 million years ago, more complex life forms, like humans (or dinosaurs), would be more vulnerable as a group than bacteria.

"You can get in this spiral of increasing complexity. And thatís what biological systems do. They tend to spiral in complexity, because you sort of keep adding new features that give you new robustness, and that creates new vulnerabilities," Dr. Doyle says. "And then a big meteor hits and you start all over."

Diana Steele is a free-lance writer based in San Diego.

© 2000 Diana Steele

Researchers are struggling with how best to design ecological preserves to protect endangered species. The Florida Panther is one of the most critically endangered species in the world.
Caltech's Dr. John Doyle says robustness always involves trade-offs -- increasing complexity adds new vulnerabilities. 

"If you understand those trade-offs, you can design them in such a way that it all works out," he says.

Diana Steele's Home Page | Science Writing | Previous Story | Next Story