© SALIM MADJD 2011 FOR ASTHMAMD
A woman brought her asthmatic son into pediatrician Sam Pejham’s Pleasanton, California, office last July because he was always tired, wasn’t sleeping, and was gaining weight. Though it was the first time that Pejham had seen the patient, he knew the boy’s asthma wasn’t under control the moment he saw the display on the mother’s smartphone.
Pejham, a researcher at the University of California, San Francisco, was looking at a line graph of a week’s worth of peak-flow data, a simple measure of lung function, using an iPhone app he had developed called AsthmaMD. Patients measure their peak flow by blowing into a separate device, then enter that data manually on the screen of their iPhone, where the data are stored and displayed graphically. Though the boy hadn’t complained of breathing problems, the graph clearly showed that his lung capacity was impaired. So Pejham prescribed a twice-daily dose of an anti-inflammatory drug in addition to the inhaler he was already using.
Health mobile data that is gathered this way is really unprecedented.
-Sam Pejham, University of California, San Francisco
“A week later, [he] was a new child,” Pejham recalls. “Now he’s able to play sports and soccer. He just never knew he could do this because he wasn’t controlling his asthma.”
AsthmaMD also allows patients to monitor their lung function themselves, and send the information to their doctors instead of making a trip to the clinic. But to basic scientists, the true value of AsthmaMD lies in the real-time data the program collects on patients’ peak flow and the geographical location where the measurement was taken, says Salim Madjd, an entrepreneur who worked on the technical side of the app. Combining those details from thousands of users (all anonymous) with the dozens of real-time data sources that track pollution levels and weather patterns, epidemiologists may soon be able to glean hourly, block-by-block information on asthma triggers. For example, they had planned to combine AsthmaMD symptom data from British users with weather pattern data to predict how the ash cloud from last year’s eruption of the Icelandic volcano Eyjafjallajökull would affect those users, but luckily the ash cloud passed them by.
And while traditional studies may have a few dozen, several hundred, or even several thousand subjects, AsthmaMD tracks tens of thousands of users. “Health mobile data that is gathered this way is really unprecedented,” Pejham says.
The app is just one of many consumer technologies that are being adapted to improve medical care, clinical trials, and basic research. New technology has always fueled development of novel drugs and better medical treatments, but historically, many of these technologies were designed by researchers working in a lab or a hospital, and were tailor-made for the job at hand. Recent advances in electronics, software, and cloud computing, however, have yielded unexpected fruits for the biomedical community. Now, consumer technologies that allow people to look up restaurant recommendations, store their photos, and keep in touch with their friends are being modified to develop drugs, test them in clinical trials, and set new standards of care in medical facilities.
Mining the data deluge
In addition to collecting data, consumer technologies can also help make sense of it. Some of the same methods that work behind the scenes to analyze stock data, for example, are also allowing doctors to integrate the deluge of physiological data that hospitals track in newborn babies, including heart rate, blood pressure, and temperature. A better understanding of how these factors change with infection could help treat newborns and even save lives, says Carolyn McGregor, a health informatics professor at the University of Ontario Institute of Technology.
Currently, doctors typically only respond to very short-term changes—for instance, when a baby has stopped breathing for too long, or if her heart rate drops too low—when deciding what treatment a newborn needs. But research has demonstrated that longer-term measurements of multiple factors, such as the frequency of apnea spells combined with heart-rate variability, may be better predictors of illness and infection. Without being able to track these more nuanced data, doctors may not always get a sense of the larger picture until the baby is quite ill, McGregor says.
To address this problem, she and her colleagues have developed a program called Artemis, which is being studied in the neonatal intensive care unit (NICU) of Toronto’s Hospital for Sick Children, as well as at a Providence, Rhode Island, hospital and a few locations in China. Unlike the routine procedure in NICUs, which overwrites a premature infant’s physiological data every few days, Artemis stores long-term data on heart rate, oxygen saturation, blood pressure, temperature, and breathing rate, and then uses a series of algorithms to look for overarching trends. By mining months of such physiological data and tying it to patients’ medical histories, the Toronto team hopes to detect significant changes in condition earlier and even uncover new predictors of illness, McGregor says. “We’re trying to let the data speak for itself.”
Print this drug
Basic drug discovery is also getting a boost from consumer technologies—specifically, from a 1980s-era printer. Titration, the serial dilution of a candidate drug for testing to determine the best dosage, is a staple of much basic lab work, but titration by hand is incredibly tedious and prone to error.
By using modified ink-jet printer cartridges that precisely inject picoliters of liquid, the HP digital dispensing system performs the titration process in essentially the opposite order. Instead of starting with a strong concentration of a compound and diluting it, the system adds cell-size droplets of active compound into a fluid to slowly strengthen the drug’s concentration, says Kevin Peters, a senior scientist with HP. The new device, which resembles a desktop inkjet printer, titrates more quickly and accurately than humans.
SIGA Technologies, a biotech company developing antiviral drugs to treat Ebola hemorrhagic fever, smallpox (eradicated, but still considered a bioterror threat), and dengue fever, uses the digital titration machine to find drug concentrations that elicit the best antiviral response in cell-culture assays. Though the company uses a sophisticated liquid-handling robot to do the initial drug screens, digital titration has made it easier to determine the best dosing for drugs further along in development, says Robert Jordan, SIGA’s director of antiviral research and development. Something that would take a day of computer programming for a researcher to set up on the liquid handler “on the HP would literally take 10 minutes,” he says.
Up in the clouds
Services such as Amazon’s Cloud Drive or Apple’s iCloud target consumers who want to listen to their music collections miles from their computers. But cloud computing, which allows users to plug into thousands of remote servers via a digital network, is also speeding up basic science and allowing clinical trials to go off the grid. (See “Harnessing the Cloud,” The Scientist, October 2010.)
At ETH Zürich, a science and technology university in Switzerland, researchers are using cloud computing to determine what makes some strains of Streptococcus bacteria so nasty. Gene sequencing has identified hundreds of virulent mutations in scores of strains, but many of them are probably never expressed, says Lars Malmström, a computational systems biologist at ETH Zürich.
To whittle down that list to the most relevant virulence factors, Malmström’s team runs millions of protein-folding simulations to predict a protein’s structure based on billions of calculations of bond energy levels in the molecule. The massive number of simulations can take a huge chunk of computing, which the researchers had traditionally done using Brutus, the school’s 10,000 CPU-core cluster. But without exclusive access to the machine, there were rarely more than 50 to 100 CPU cores available at a time, Malmström says, and the team would have to ask the university’s help desk for approval to tweak every little setting in the program.
To speed up the process, the team enlisted the help of SmartCloud Enterprise, IBM’s cloud for commercial users, in partnership with CloudBroker, a high-performance cloud computing company. It allows Malmström and his colleagues to draw on the computing power of CPUs around the world without having to install specialized software required by other distributed computing methods. Using the program, they were able to narrow down the list of virulence factors in just two weeks. The task would have taken the group months using Brutus.
And unlike servers sitting in their lab, the researchers only pay for the service when they’re actually running it. “You can have 1,000 computers when you need them, and have zero computers when you don’t need them,” Malmström says.
Cloud computing is also improving clinical trial management in remote locations. The Aeras Global TB Vaccine Foundation, a nonprofit organization that has three ongoing tuberculosis vaccine trials running in South Africa and Kenya, is using Cmed Technology’s cloud program called Timaeus to monitor patient data in real time. Doctors or nurses upload data to private US servers via cell-phone networks, which are more reliable than Internet connections in Africa, so trial managers around the world can always access the most recent data.
“I can log into the system and check up on the status of certain data points of interest any time I’d like,” says Margaret Ann Snowden, the biostatistics and data manager at Aeras. And if there’s a mistake, she doesn’t have to ship data-clarification forms out to remote sites and wait for them to ship back the forms; she simply submits the questions online. The technology has also made the process of patient information intake quicker and more accurate, she adds, because the system automatically catches small errors in the data, such as date transposition, and eliminates the need to decipher handwritten charts. And like IBM’s SmartCloud, it’s also flexible. If recruitment is going slowly, for instance, Aeras only pays data-management fees for the number of patients actually enrolled.
How all these technologies will alter the biotech landscape is still an open question, but Madjd likens the difference to the world before and after the Internet revolution. E-mail didn’t necessarily change how people communicated, “it just made it faster and quicker,” Madjd says. “And that’s true for doing medicine using these new technologies.”