by Anne Stuart
What might it mean for science and medicine if researchers could analyze biochemical
data faster and more accurately?
On a personal level, what if you could track your exercise level all day long,
then decide after work whether you still needed to log an hour at the gym?
Or what if you could monitor your vital signs round-the-clock, giving your
physician an up-to-the-minute portrait of your body’s workings instead
of the snapshot provided at your yearly physicals?
Those are just a few of the scenarios that could result from the latest research
being conducted by HP Labs.
For the past year, HP Labs scientists in Cambridge, MA, have
been applying their expertise in such diverse areas as signal
analysis, high-performance scalable systems, wireless mobile devices,
machine learning, data search and indexing to the life sciences
domain.
Their ultimate goal: "To improve the overall quality of
health care while lowering the cost," says researcher Beth
Logan.
The ambitious mission is a new one for the group, whose previous
projects include SpeechBot
, a search engine that finds and indexes
spoken audio on the Internet.
"We've been changing our direction to health and
life sciences because there's so much growth in that area,"
says Logan as she and her fellow researchers -- Amir Bar-Or, Scott
Blackwell, Dave Goddeau, Jennifer Healey, Leonidas Kontothanassis,
Alex Nelson, and J.M. Van Thong -- settle into a cluster
of sofas and beanbag chairs in their twelfth-floor meeting space.
They don't need to look far for evidence of that growth.
Before them, a low table overflows with research papers, journals,
and magazines with names like Bio-It World and Circulation. Downstairs
and around the corner is the Massachusetts Institute of Technology,
a hotbed of bioscientific research; the headquarters for some of
the world's leading biotech firms are within a brisk walk.
The group's current and planned research stems from four
major factors affecting wellness:
- Genomics, which examines the molecular basis for disease
and medical treatment.
- Environment, which explores how genomic
factors in the air and water influence human health.
- Lifestyle,
which emphasizes methods people can use to personally monitor
the impact of their daily lives on their overall wellness.
- Health care, which focuses on ways to provide physicians with
patients' comprehensive
real-time physiological data and other health-care providers
with mobile access to medical information and resources.
"These factors contribute equally to an individual's
wellness, yet spending is totally focused on health care. The focus
is on treatment of illness, rather than its prevention and early
detection," says Richard Zippel, director of HP Labs' Cambridge
Research Lab and technical lead for HP's efforts in wellness. (See
a related video).
HP is already the leading supplier of computing systems for
life sciences research -- the human genome was sequenced on
HP computing systems -- but Zippel thinks the company can contribute
much more, thanks to its expertise in areas like technical computing,
mobility and database management.
"HP can have an effect on all aspects of wellness, so instead
of being reactive to diseases, medicine becomes proactive," he
says.
Zippel's team doesn't include any medical experts
-- researchers joke about learning medicine by watching the television
show "ER" -- but partners with other organizations
that contribute in-depth knowledge about the scientific areas under
study. The list of collaborators includes such world-class institutions
as Massachusetts General Hospital and Brigham and Womens Hospital,
the Harvard Partners Center for Genetics and Genomics and The Center
for the Advancement of Genomics.
In one such joint genomics project, HP Labs scientists are working with Boston
University to develop better ways to classify proteins.
"Although scientists
have obtained the sequence information for a growing number of organisms,
we do not know the function of many of the proteins described by these sequences,"
Logan explains. "The point of our project is to get some idea of these
functions by predicting protein structure," research that
has long-term implications for the development of new disease-fighting drugs.
Scientists know the sequences of hundreds of thousands of proteins, but so
far have identified the structures of only a few thousand. That's because
these structures are so complex that analyzing them manually through biological
experiments is a labor-intensive, time-consuming process. Using its machine-learning
expertise, the group developed techniques for automatically classifying proteins
by comparing them to a database of knowledge about known structures.
Students at Boston University have been helping to test the procedure by attempting
to classify 300 to 400 proteins. Ultimately, the partners plan to establish
a university-based Web site where anyone can upload a protein sequence for
comparison with hundreds of known structures, with results returned in minutes
by e-mail.
In another collaboration, HP has been working with Harvard
Partners on a mass spectrometry project that determines the protein content of biological
samples such as blood, saliva, and urine. Currently, that is a cumbersome,
error-prone procedure.
"The process is so inexact that the spectrometer product is junk two-thirds
of the time," says HP Labs researcher Kontothanassis.
Again using machine-learning
tools, the group developed techniques for identifying useless results,
eliminating them and speeding up the spectrometry process by a factor of
two. Researchers sped up the process even more by searching the database
of protein spectra in an intelligent manner.
Ultimately, the group hopes to use mass spectrometry to quickly and accurately
identify "biomarkers" for disease in biological samples.
"Sick people and well people may have different concentrations of
proteins in their blood, for example," Logan says. "The hope
is that if we can refine this technique, we'll be able to use it for
diagnosis."
On the health and lifestyle front, the lab is experimenting with methods
for monitoring human physiology, typically storing information for later
analysis. Using their signal-analysis and system architecture background,
the researchers on the BioStream Project are building an interface that can
collect patient physiological data (primarily electrocardiogram signals,
blood pressure, temperature and weight) and analyze it for indications of
problems or illness.
Researchers see multiple applications for that capability. Hospital emergency-room
teams might use the data for triage, determining which incoming patients
need treatment first -- a natural extension of HP's existing solutions
for hospitals. Individuals might constantly monitor their activity levels,
making health and lifestyle choices accordingly -- deciding, for instance,
whether to end the day with a strenuous workout or just a brisk walk.
A third application would be a home monitor for patients with actual or
suspected heart problems. A five-minute electrocardiogram in a doctor's
office "is a snapshot of your heart," says HP researcher Healey. "Based
on that, doctors must make a diagnosis. Sometimes they can, but some diseases
require longer to detect."
In conjunction with others at Hewlett Packard, the wellness researchers
are developing a system that could monitor patients at home for weeks or
months at a time, allowing physicians to pinpoint problems earlier.
"Our vision is that in the future you will be monitoring all your
vital functions 24 x 7," says Van Thong. "Once you are able to gather
these physical signals for a long period of time, you would be able to do
preventative health care, or respond more quickly when there's a problem."
While the group's work encompasses a diverse, complex, and growing
range of life-science initiatives, all of their efforts are focused on
a single desired outcome: helping people live longer, healthier lives.
|
Printable version
