Editorial: Rule out nothing in the investigation of cancer
WHEN Danny Hillis spent a day watching a top surgeon perform keyhole cancer surgery, he was left both exhilarated and depressed. The clinical precision with which the surgeon opened up the patient, used state-of-the-art robotic tools to remove their tumour, and sewed them back up again was breathtaking. It was also deeply disheartening. 鈥淲ith all our science, the best we can do is try to cut the cancer out with a knife,鈥� says Hillis. 鈥淭hat is the caveman approach to disease.鈥�
A few years ago, what he thought would not have mattered. who pioneered the concept of massively parallel computing. His accomplishments include , now exhibited at the Museum of Science in Boston, and creating cutting-edge computer systems for Walt Disney theme park rides and animations. Impressive, but hardly a convincing resum茅 to pronounce on the shortcomings of modern cancer surgery.
Advertisement
鈥淗illis had designed theme-park rides for Disney 鈥� hardly a convincing resum茅 to pronounce on modern cancer surgery鈥�
But Hillis鈥檚 thoughts and expertise do matter. According to the World Health Organization, . While traditional biological approaches to understanding and combating cancer have had some great successes, mortality rates remain stubbornly high. That鈥檚 why last year the US (NCI) enlisted Hillis and other high-powered researchers from physics, engineering, mathematics and computer science to see what extra ammunition they might supply.
The newcomers will not be required to design new death rays or high-powered imaging devices for tumours, the traditional metier of physical scientists. Instead, they will harness the mathematical tools and the broad-brush approaches of their disciplines for a new goal: to find simple laws that describe a cancer cell鈥檚 fate as surely as they do a falling apple.
That is a world away from the traditional, 鈥渂ottom up鈥� approach to looking at cancer. For the past few decades, research into the disease has focused on its molecular basis: identifying the genetic mutations and protein abnormalities that underlie it, and designing drugs to treat them. The US Food and Drug Administration has approved something like stemming from this approach. Coupled with earlier diagnoses, the result has been a decline in death rates from almost all cancers in wealthy nations.
Yet all too often those advances have been incremental. In the US, for example, the five-year survival rate for pancreatic cancer increased from 3 per cent in 1975 to just 5 per cent in 2004. 鈥淲e have a lot of drugs, but they鈥檙e not doing that much,鈥� says , an oncologist at the University of Southern California (USC) in Los Angeles. 鈥淭hey haven鈥檛 really changed the playing field.鈥�
The trouble is that if its normal pathway to growth is blocked by a drug, a tumour will often simply evolve and find new molecular channels through which to grow and prosper. 鈥淲e keep hammering at each individual signalling molecule, but we forget that there鈥檚 this whole complex system that it is interacting with,鈥� says Agus.
It was Agus who invited Hillis to see cancer surgery at first hand. Bizarrely, the impetus for their collaboration came from former US vice-president Al Gore. In 2003 Gore visited the centre in Los Angeles where Agus was pioneering techniques to understand how communication between proteins affects the onset and progression of cancer. The project was producing huge amounts of data, and Gore suggested it might benefit from specialist number-crunching expertise. 鈥淗e said I could really benefit from getting an engineer or a mathematician involved, and suggested Hillis,鈥� says Agus. 鈥淚 thought, do I really want to meet a guy from Disney who builds computers?鈥�
Initially, Hillis was no less sceptical. But the two agreed to meet and soon found they had a lot to talk about. As a systems engineer, Hillis was puzzled by the way oncologists reduce cancer to the action of genes, when it is a disease that affects a whole organism. Agus too found himself looking at cancer in a different way. 鈥淭he natural inclination of physical scientists is to step back and say, 鈥楲isten, I know I鈥檓 not going to understand every component of that system. But it doesn鈥檛 mean I can鈥檛 control that system鈥�,鈥� he says.
Since last year, their collaboration has been formalised within the new at USC, one of 12 that the NCI has set up across the US. Agus and Hillis are now working together to create a model of how lymphoma, a cancer of the immune system, grows and spreads in mice. The project showcases the sophisticated technologies physicists and engineers already contribute to cancer research: nanoscale sensors for analysing protein interactions, microfluidic tools for looking at genetic changes in single cells and even microscopic techniques that allow a direct peek at tumours growing inside living organisms.
The real innovation, though, is what Agus, Hillis and around 20 other scientists from disparate disciplines are doing with all the data streaming in from that research. They are testing a set of interlocking computational models they have developed from basic principles to describe and predict different aspects of the disease: from protein interactions and modifications within cells, through a tumour鈥檚 growth and genetic evolution, to the host鈥檚 response to the disease and various therapies.
Personalised therapy
Within five years, the team hopes to have a single, all-embracing model of mouse lymphoma that fits the data as convincingly, and has as much predictive power, as a theory of gravity or electric fields. If they can pull that off, transferring the insights to humans will be the next challenge. The dream is to produce a model that, by plugging in key parameters 鈥� sex, blood pressure, genetic sequences and the like 鈥� could predict an individual鈥檚 response to various combinations of cancer therapies. 鈥淲e could run a simulation for each patient, and design a complex treatment specific to that patient,鈥� says Hillis. 鈥淢aybe we would use some radiation for while, then heat them up, then change the glucose level in their blood.鈥�
That would be an extraordinary turnaround for cancer therapy. 鈥淭he science behind what we do every day as oncologists is almost 50 years old,鈥� says of the Westside Prostate Cancer Center in Beverly Hills, California, who is also involved in the USC centre. The experiments that form the scientific basis for giving chemotherapy and radiation were done in the mid-1950s, he says, and since then optimisation has been largely about coming up with less toxic and more effective chemicals, or using a different schedule of treatment. 鈥淧atients ask me, 鈥榃hy are you giving me six doses of this chemotherapy instead of four or eight?鈥�,鈥� says Gross. 鈥淭he reason is, the other thousand patients got six doses of this chemotherapy, so we have to give you the same thing.鈥�
Not everybody is convinced by the new approach. Many biologists doubt the human body will give up its mysteries under crude mathematical scrutiny. Such an approach implies a degree of simplification and loss of detail that could never do justice to the complexities of a dynamic and variable system, they argue. 鈥淵ou can come up with computational models, and they may be very interesting, but the question is whether they reflect reality,鈥� says , an oncologist at the Albert Einstein College of Medicine of Yeshiva University in New York.
of Arizona State University in Tempe, a distinguished theoretical physicist and cosmologist recently recruited by the NCI to its new programme, takes a different view. 鈥淓very system around us is in practice very complicated,鈥� he says. Armed with basic rules of nuclear physics, magnetic fields and heat radiation, for example, we can produce pretty good models of how the sun works, without knowing the details of what every particle inside it is doing. Why should the same not be true for living organisms? Augenlicht counters that whereas stars such as the sun develop according to a well-defined pattern, biological entities evolve in response to their environments in ways that might not seem inherently logical or efficient. 鈥淭rying to derive a model of how cells behave, how tumours behave, is very different to working out how the universe behaves,鈥� he says.
鈥淲e can model how a star works. Why should the same not be true for living organisms?鈥�
is one researcher taking up the challenge. He cut his teeth in USC鈥檚 aerospace and mechanical engineering department developing numerical models of hurricanes. As a project leader at the new physical sciences oncology centre at the Scripps Research Institute in La Jolla, California, he is now also applying his modelling expertise to metastasis 鈥� the poorly understood and lethal process by which tumour cells break away from their original site and spread via the lymphatic and blood circulatory systems.
Break it down
Newton aims to tackle metastasis as a physicist or engineer would: by breaking it down into simple steps that can each be modelled using equations. 鈥淲e write down the fundamental equations of fluid mechanics for the blood flow and for a deformable cell, model what a deformable cell is, model the vessel walls the blood is going through, model the pumping of the heart and the pulsatory fluid motion that鈥檚 driving the whole system, and then numerically simulate it,鈥� he says. 鈥淚n principle, the equations are all there.鈥�
What he needs, and is getting, are the specifics: basic stuff such as how stiff cancer cells are and how far they can stretch and squish without breaking. In his centre in Tempe, similar questions are on Davies鈥檚 mind. 鈥淚t鈥檚 quite clear that cancer cells are physically distinct from healthy, and as they become more and more malignant those changes become more and more pronounced,鈥� he says. His team is looking closer at those changes, using the technique of atomic force microscopy to measure the shape and stiffness of cancer cells and how the nanoscale structure of chromosomes changes as the disease progresses.
No one knows how informative this concentration on the mechanical properties of cells rather than its genetics will turn out to be 鈥� perhaps not very, says , director of the NCI鈥檚 physical sciences initiative. 鈥淏ut if it is, it will open up a new field.鈥�
鈥淪ome people definitely feel that the approach is a little abstract, or that we鈥檙e oversimplifying things to the point where they鈥檙e irrelevant,鈥� says Parag Mallick, a proteomics researcher at the USC centre. , an oncologist at the Stanford Cancer Center in California, welcomes any fresh approach, but questions the likelihood of quick insights. 鈥淚t鈥檚 exciting and new, something we haven鈥檛 tried before,鈥� she says. 鈥淏ut the speed at which this develops into something useful in diagnosing and treating cancers, that鈥檚 still up in the air.鈥�
Those involved stress that it is about learning from each other, not about physicists taking over and showing biologists everything they are doing wrong. 鈥淲e don鈥檛 have the knowledge of the biological system,鈥� says Hillis, the computer scientist. 鈥淭here鈥檚 no way we could do this ourselves.鈥�
To Agus the oncologist, meanwhile, that knowledge gap might turn out to be an advantage: great insights often come from those unencumbered by intellectual baggage. 鈥淕alileo would go out every night and map stars,鈥� he says. 鈥淎fter four months he had a beautiful map where he could predict where every star would be. But he didn鈥檛 even know what a star was.鈥�
