August 2003
http://www.coe.berkeley.edu/labnotes/0803
Commuter-friendly version

Professor Teresa Head-Gordon cross-disciplinary research combines experimental, theoretical, and computational approaches. Angela Privin photo

In Teresa Head-Gordon's laboratory, an IBM supercomputer cranks out dreamlike visualizations that are reminiscent of artist Salvador Dali's surreal landscapes. But these stunning graphics are not eye-candy, they're precise representations of proteins, the building blocks of human life. Ultimately, the bioengineering professor's mind-boggling models could lead to cures for diseases like Parkinson's and Alzheimer's.

Head-Gordon's efforts are focused on simplifying protein models so that scientists can more easily understand how proteins form and potentially alter disease-causing characteristics.

Proteins consist of long chains of 20 kinds of amino acids linked together under instruction from DNA. Once the chain is complete, the protein literally folds itself up so that only a portion of the resulting molecule is exposed to react with the environment around it. For example, the shape of the protein hemoglobin makes it ideal for the transportation of oxygen through the bloodstream while collagen is perfectly structured to form connective tissue.

The detailed protein folding model (top) above is much more complex and difficult to analyze than Head-Gordon's minimalist model (below) of the same molecule. Image courtesy the researcher

One unusual property of proteins is that even though their concentration in the body may be incredibly dense, they avoid clumping together, a phenomena known as aggregation. Protein aggregation around nerve cells is associated with the debilitating symptoms of Parkinson's, Alzheimer's, and other diseases.

"There's something specific in the sequence of amino acids that essentially can result in aggregation," Head-Gordon says. "So if we can understand those molecular processes at the sequence level then in principle we could potentially re-engineer these disease proteins to avoid aggregation."

Head-Gordon's innovation lies in her minimalist protein models. Think of a protein as a necklace beaded with various amino acids. Rather than model each bead— a computationally-difficult and time consuming process —Head-Gordon categorizes the beads into just three flavors based on their physical interactions during folding. While this coarseness of detail does not depict every component of a protein, it represents enough to provide scientists with insight into what combinations of beads may result in aggregation.

After Head-Gordon generates a model, her collaborator in the Department of Chemical Engineering, professor Harvey Blanch, verifies in vitro whether the engineered proteins aggregate as predicted by the simulation.

"The reduced computational cost at the coarse-grained level of abstraction will potentially enable both folding studies on a genomic scale and systematic application in protein design," Head-Gordon and postdoctoral researcher Scott Brown write in a recent scientific paper.

While the minimalist models could someday lead to gene therapies for certain diseases, Head-Gordon says those health-related applications are not in the near-term. More immediate, she explains, are benefits for biotechnology companies that produce proteins for pharmaceutical use and research. Often, the synthesis machinery in the bacteria that the biotech industry uses to produce proteins becomes overwhelmed. This results in the over-expression of the protein and, ultimately, protein aggregation. A denaturing step is then required to reverse the aggregation. Engineering the proteins not to aggregate would reduce the cost of this intermediate step and enable the company to produce more proteins in a shorter period of time.

From today's biotech industry to tomorrow's gene therapies, Head-Gordon's minimalist models are proving that less is indeed more.

"Sometimes when you have so much detail, you get lost in the forest," Head-Gordon says. "With minimalist models, things are much easier to characterize, analyze, and understand."

Berkeley computer science professor David Culler is the director of the Intel Research Berkeley Laboratory and is also a principal investigator on the PlanetLab project. Peg Skorpinski photo

Imagine you've arrived in Paris for a conference and you sit down at an Internet station in the hotel lobby. You're thousands of miles from your office in Berkeley, California, but your familiar computer desktop instantly bursts onto the screen. Any data you need is fetched instantly and your most processor-hungry applications run without a hitch. No fuss, no muss, and most impressively, no lag. Somehow, all of your data and computing power has followed you across the world.

This next-generation Internet application is just one of the geographically-distributed online services currently in development on PlanetLab, a global online test bed developed by UC Berkeley, Intel Research, Hewlett Packard, and other university collaborators around the world.

First conceived of by Berkeley computer science professor David Culler, the director of the Intel Research Berkeley Laboratory, and Princeton University professor Larry Peterson, PlanetLab is now running on 170 computers at 60 research centers around the world. In the next few years, Culler expects the network to grow to more than 1,000 computers installed in a diverse array of academic institutions, corporations, Internet service providers and homes.

Unlike most of today's Internet services and applications that are based at one Web site, PlanetLab enables pieces of new applications to run on many computers around the world, self-organizing to form their own networks and enabling processing to occur inside the network instead of at its edges.

"Increasingly important services are going to be implemented as capability spreads over much of the Internet instead of being concentrated at a few points," says Culler, whose work is part of the Center for Information Technology Research in the Interest of Society (CITRIS).

For example, an overlay network like PlanetLab could lead to more robust video multicasting. Today, a Web site hosting a popular video clip can become overloaded with too many requests from viewers. An overlay network could automatically redirect requests for the video to sites hosting the same content that are nearer to the viewer, eliminating traffic jams and increasing download speeds for the user. PlanetLab may also lead to new methods to protect the Internet from viruses and worms and the development of "persistent" storage capabilities that give the Internet a "memory."

This kind of online "storage utility" is the aim of OceanStore, one of the UC Berkeley projects experimentally deployed on PlanetLab. Led by computer science professor John Kubiatowicz, OceanStore is essentially a massively-distributed hard drive system. The system copies and spreads a user's data to servers around the world for safekeeping and quick access.

"One of the motivations for developing PlanetLab was that there were leading researchers at Berkeley who had a clear direction in their projects, such as OceanStore, but no test-bed for trying it out," Culler says.

Professors Paul Wright (pictured) and Ed Arens are collaborating with PhD candidates Nathan Ota (ME) and Therese Pfeffer (Architecture) on the demand response energy system research. Bart Nagel photo

As the summer temperature in California rises, so does the risk of brown outs. A spike in demand combined with the state's energy crisis means higher utility bills. To dramatically cut the cost of keeping cool, UC Berkeley researchers are developing a consumption-aware, cost-saving technology combining "demand response" energy pricing with a network of tiny sensors and smart thermostats for the home.

"From June to September, there are huge peaks in our energy demand, " says mechanical engineering professor Paul Wright, a principal investigator on the project. "Air conditioning accounts for up to a 50 percent increase over baseline consumption. Wouldn't it be great if your thermostat and meter could receive information about when the price is lowest to run your air conditioner and adjust your thermostat to reflect that information?"

Wright, Edward Arens, the director of UC Berkeley's Center for the Built Environment (CBE), and Cliff Federspiel, a researcher at the CBE, are building such a system with support from the California Energy Commission. The three are collaborating on the effort with professors David Auslander, Jan Rabaey, and others in the Berkeley Wireless Research Center, professor Richard White of the Berkeley Sensor and Actuator Center and his students, and a group led by professor David Culler, director of the Intel Research Berkeley laboratory. The multi-disciplinary project falls under the umbrella of the Center for Information Technology in the Interest of Society (CITRIS).

The slide above shows the general wireless scenario. A new style thermostat would receive information from the utility company, allowing the consumer to adjust the usage on AC and appliances.

The technological foundation for the system is a network of tiny wireless sensors that could be easily installed throughout a home. The sensors monitor temperatures in various parts of a house and relay that data back to a central computer for processing. The Berkeley researchers are developing the wireless sensors, including an "energy scavenging" technology that converts the ambient vibration of structural components like air-conditioning ducts into electricity.

"The meter needs to know precisely when you use your electricity in order for the utility company to bill you based on the time of day," Wright says.

Meanwhile, as energy prices shift throughout the day they would be transmitted wirelessly from the utility company to a smart meter at the home. The resident's only responsibility would be to program this or her temperature preferences on a user-friendly smart thermostat. Employing new control algorithms designed by the researchers, the system could then set the air conditioning system to match the desired temperature profile even as it changes throughout the day.

"Even if you're home during the day, there are ways to spread the thermal mass load of the house to keep it cool without turning on the air conditioner at peak times when the energy is most expensive," Wright says.

Of course, the key to such a system is for the energy companies to create a demand response pricing structure for households. Wright is confident that once the "client-side" technology is proven, the utilities will oblige.

"Big commercial buildings in the Bay Area already have time-of-use pricing on their meters," he says. "But the goal of this project is to diffuse it all the way into California homes."

Video Highlight Researchers working with the Center for Information Technology Research in the Interest of Society (CITRIS) talk about how CITRIS research can change our world. Both videos are Windows Media Short Version (3:15) Full-Length Version (6:05)

In 1999, a profound vision emerged from UC Berkeley's College of Engineering. It was a decidedly grand challenge: Create a multi-disciplinary center where researchers could collaboratively develop information technology to tackle society's biggest problems.

"We need to re-engineer engineering in the context of today's world," one faculty member said at the time."

Driven by Dean Richard Newton, CITRIS was formally founded on July 1, 2001 by UC Berkeley, Davis, Merced, and Santa Cruz as one of four California Institutes of Science and Innovation established by Governor Gray Davis. Initial funding of $20 million in the 2001-2002 state budget, combined with corporate and private pledges of more than $170 million, followed shortly after. But even before the money was pledged, the research had begun.

When Ruzena Bajcsy, the former Director of Computer Information Science and Engineering at the National Science Foundation, took her post as CITRIS's first director in October of 2001, there were 83 participating faculty. Today, there are more than 200 CITRIS researchers from more than 50 departments across all four campuses. Groundbreaking for an 80,000 square-foot CITRIS building, including a new state-of-the-art microelectronics/nanofabrication facility, is slated to take place in the spring of 2004. Gary Baldwin, who previously directed the Gigascale Silicon Research Center, is the CITRIS executive director while Berkeley computer science professor James Demmel, as the CITRIS chief scientist, helps coordinate the 150+ research efforts under the CITRIS umbrella.

Corresponding to the CITRIS proposal and mission statement, the center's research is divided into seven categories: energy efficiency; transportation; emergency response and homeland defense; education; environmental monitoring and management; health care; and social sciences, humanities, and business. From fighting the state's energy crisis using CITRIS-developed networks of Smart Dust sensors to exporting Berkeley's computer science curriculum through new distance learning technologies to monitoring buildings and bridges for structural integrity after an earthquake, CITRIS has become a thriving hub of innovation in just a few short years.

Most recently, CITRIS researchers landed a $1.65 million grant by the California Energy Commission to develop demand-responsive "smart" thermostats. This summer, CITRIS researchers collaborated with the Intel Research Berkeley lablet and biologists to deploy a novel environmental monitoring system on a small island off the coast of Maine. Meanwhile, the Chicago Fire Department is collaborating with CITRIS to develop a system of wireless sensors and transceivers that enable firefighters to navigate more safely through a burning building.

And that's just the beginning.

"With each new milestone," Bajcsy says, "we are laying the foundation needed to fulfill the goals of our original charter—to sponsor collaborative information technology research that will ultimately provide solutions to grand-challenge social and commercial problems affecting the quality of life of individuals and organizations."

Original article: If You Can See This, You're Too Close http://www.coe.berkeley.edu/labnotes/0502/lightbar.html An innovative signal light developed at UC Berkeley that warns drivers when to back off the tail of a bus is currently being tested on two buses in Ann Arbor, Michigan. Designed by Berkeley professor of vision science and bioengineering Theodore E. Cohn, the system consists of a five-foot wide bar of LED lights mounted on the rear of the bus. When a radar system also mounted on the back of the bus detects that a car is too close, approaching too rapidly, or both, it triggers the light bar warning system. The lights then flash in a pattern optimized to take advantage of the fastest pathways in a human's visual nervous system. The Ann Arbor Transit Administration is conducting the field tests with research consultants Veridian Inc. The research was first profiled in the May/June 2002 issue of Lab Notes. Articles in the New York Times and Chicago Tribune followed.

Comments, questions, suggestions?
Send us your feedback by emailing lab-notes@coe.berkeley.edu.

June/July 2003

A Force Field for No-Fly Zones Speaking as a pilot, the idea is outlandish. To be at all useful, such a system would necessarily be impossible for a pilot to disable or bypass. I challenge you to name any existing aircraft electronic system of even one-tenth the complexity required for "soft walls" that has not failed and needed to be disabled for flight safety. You're trying to solve a problem that has already solved itself. No one will ever again successfully take over an airliner. Before 9-11, the best move was for passengers to cooperate. That will never again be the case. — Keith Burton - PP, ASEL; Wyncote, PA Response from Edward Lee Pilots frequently react this way to the Softwalls proposal. A pilot comes from a 2000-year-old tradition of the "captain of the ship," where even the authority to marry the passengers is granted. The captain is responsible for the ship, its crew, and its passengers, and the tradition dictates absolute control over all elements of the craft. Presumably, this is why transponders in commercial aircraft have an "off" switch in the cockpit. The Sept. 11 hijackers used this "feature" to delay detection of their intentions. In retrospect, it is clearly unconscionable to grant pilots this authority. The risk of a fatal malfunction in the transponder is so small compared to the damage done by turning it off that the safety of the people on the ground trumps the pilot's authority. The fact that equipment fails is real. The skepticism about new devices in aircraft is healthy. But the fact is that aircraft have consistently gotten both more complex and safer. This particular pilot probably should never assume control of Boeing 777, since the fly-by-wire electronics removes all mechanical couplings so that if the cockpit electronics fail, then the craft will crash. This pilot claims that the problem has solved itself. Apparently, the Pentagon does not agree, since critical sites in Washington DC are now protected by antiaircraft batteries. In my opinion, the mere presence of this protection scheme poses a far greater risk to pilots, crew, and passengers than that posed by the risk of failure of well-designed cockpit electronics (witness the impressive safety record of the 777). You asked for my opinion on "soft walls", giving me the choice between "science fiction" and "homeland security". Given those choices, I'd have to call it science fiction. I've even *seen* it in science fiction (specifically _Stations of the Tide_, by Michael Swanwick, 1991, ISBN 0-380-71524-4, in Chapter 12, entitled "Across the Ancient Causeway"). I'm sure it is really fun to work on, and perhaps is a funding magnet in these times of national security driven research, and I know that those two things alone are sufficient. As the article said, the computer security aspects of the system are a "huge concern". Any database of no-fly-zones can be outdated, subverted, or just plain pithed. And what about "griefers", as they are called in the gaming world? Can a big jetliner be herded into a no-fly-zone by a number of small aircraft, while trying to avoid collision with them? Thanks for your time. — Dan Liddell

June/July 2003

Sharing A Vision The success of a ShareCam would seem to rest in the idea that a democratically-activated camera is more interesting than fixed view or a view that moves through a preprogrammed routine. I think a related computer experiment would be to use a higher resolution camera (3 or 4 megapixel) that provides a fixed image of a scene and then serve a 200x200 pixel feed of any portion of the image, at any requested resolution, to any number of users. The first user could request a one-to-one pixel resolution of a specific object and see their request at 72dpi in either real time or at whatever refresh rate is designated by the software. Another user could request a one-to-four resolution of the same area while another could request a view from a different area of the image entirely. You could even view the entire image at once in all its compressed glory. Each user could then "move" their camera as they see fit and feel as if they are in complete control of what they see. The question is: how could you pull multiple data requests from the same source image simultaneously and feed the results to many different recipients? Especially if the image is refreshing every second or faster. — Matt Budke Response from Dezhen Song, PhD Candidate Thanks for your feedback for ShareCam system. The high resolution camera available today can offer 3~5Mega pixeled image. The problem is those high end cameras have same horizontal field of view as low end versions. The stand HFOV with tolerable quality is about 50~60 degrees, which is not able to cover the 180 degree pan range if using only one camera. It is possible to use multiple cameras but cost and complexity of the system will increase a lot. On the other hand, it is possible to use a panorama camera to cover more than 180 degrees. But, as mentioned in Ken's email, a significant distortion will be introduced by its lens. Panorama cameras usually come in low res versions. So, it is hard to correct the distortion by mathematic transformation if the initial information is too limited.