Database May Help Diagnose Illnesses

Harvard researchers are developing computer software that they say will revolutionize medicine by sifting rapidly through vast amounts of data to diagnose illnesses and predict treatment effectiveness.

“The goal is to try to improve the translation of research from the bench to the bedside as quickly as possible,” said Scott T. Weiss, a professor at the Medical School who is involved in the venture.

The $20 million project—an interdisciplinary collaboration between computer scientists, doctors, and geneticists—is part of a National Institutes of Health (NIH) initiative to improve bioinformatics, the science of storing and analyzing biological information.

The researchers competed against hundreds of other applicants for the money, Weiss said.

They have just completed the first step of the five-year project—a pilot study focusing on asthma.

The computer is programmed to look at a patient’s medical record for factors, like smoking, that might have led them to hospitalization. This result is then compared with the doctor’s analysis.

“We’ve finished [the diagnoses of released patients] and over 97 or 98 percent of the time the expert and the computer agreed,” said Weiss, who headed the asthma team.

If further trials on ailments like diabetes, hypertension, and Huntington’s Disease prove successful, the software—public property, as specified in the grant—could have far-reaching predictive implications for patient care, said John Glaser, the project’s co-director and chief information officer of Partners HealthCare, a consortium of Boston-area hospitals.

Researchers intend to scan 2.5 million consenting-patient profiles gathered from hospitals affiliated with the Partners network.

Although such a large concentration of data could raise concerns about privacy, the information is protected by numerous passwords, firewalls, and scrambling, the doctors said.

In the future, the software could be used to develop a DNA biobank, Weiss said—although he acknowledged that this is many years and logistical hurdles away.

“I’m not clairvoyant, but it’s not inconceivable that big DNA banks could become a thing of the future,” Weiss said.

The software would mean that studies that currently take years to produce due to the time it takes for humans to gather and analyze the data could be compiled in moments. But it may also be applicable to patient care, by making it easier to identify symptoms and warn patients of genetic predispositions.

“You’ve got a situation where we could intervene and make things better,” Weiss said.

—Staff writer Natalie I. Sherman can be reached at nsherman@fas.harvard.edu.