OSU: Big data helping improve diagnoses, treatment

Patients are now being more precisely diagnosed and treated, thanks to an Oregon State University researcher’s work in translational data science.

A key to that science is ontologies: systematic descriptions of knowledge that allow for integration and analysis of lots of data, in this case medical data.

In a paper published today in the New England Journal of Medicine, Melissa Haendel of OSU’s Linus Pauling Institute describes ontologies and how computers use them to enhance physicians’ decision making.

“Huge amounts of high-throughput data, including those obtained through genomic, proteomic and metabolomic analyses, are now being used in clinical analyses,” Haendel said. “The volume and depth of data and the rate at which data are being obtained are unprecedented in human history.”

Haendel directs Oregon State’s Translational and Integrative Sciences Laboratory, which aims to apply data science principles, techniques and technologies to large-scale problem solving on a societal level.

Precision medicine is one of the challenges translational data science is taking on.

“It requires precise characterization of patients: signs, symptoms, demographics, genes and the environment,” Haendel said. “Much of that information is available but it comes from different sources in formats and structures that aren’t compatible. That’s where ontologies come in. They define relationships between concepts in a way that allows computers to do logical reasoning.”

Take, for example, a man who goes to a cardiologist about heart problems, to a sleep specialist about sleep apnea, to a podiatrist about hammer toe, and to an emergency room with a collapsed lung.

The patient is likely unaware that the various symptoms may have an underlying common cause, and if he reports a fragmented symptom profile to each doctor, any one of them might easily overlook the possibility that the patient has Marfan Syndrome, a rare genetic disease that has been proposed as the explanation for Abraham Lincoln’s lanky, long-armed stature.

More than 7,000 rare diseases have been identified, and that number is growing. While a physician may have seen many cases of diabetes, she may have only seen a single case of a particular rare disease in her entire career.

Ontology-driven representations of medical knowledge can help guide physicians to ask the most relevant questions and draw clearer connections between symptoms and diseases, thereby arriving more quickly and reliably at the best course of action.

Leveraging ontologies to unify the large amount of research data, from mice to monkeys, makes logical leaps possible – from other organisms to humans and from lab bench to bedside.

Clinical judgment will always remain central to medical decision making, Haendel notes. By unburdening physicians from routine information-retrieval tasks, however, and providing a computational foundation for the interpretation of medical big data, ontologies pave the way for the use of genomic and other high-throughput data to arrive at the best decisions.

Collaborating with Haendel on the paper were Christopher Chute of Johns Hopkins University and Peter Robinson of the University of Connecticut.

The National Institutes of Health and the National Center for Advancing Translational Sciences provided funding.

About the Linus Pauling Institute: The Linus Pauling Institute at OSU is a world leader in the study of micronutrients and their role in promoting optimum health or preventing and treating disease. Major areas of research include heart disease, cancer, aging and neurodegenerative disease.