NUMBER FOUR: MAKING LUPUS AND MS MORE PREDICTABLE

Autoimmune disorders, like multiple sclerosis and lupus, are notorious for their unpredictability. The illness flares—often without any obvious trigger—and then recedes, and then flares again. Some patients deteriorate rapidly, while others can live for years before the most severe symptoms arrive. This fog of unpredictability can make it very difficult for doctors and patients to choose treatment plans.

Suchi Saria, an assistant professor of computer science at the Whiting School, believes she can shine a light through this fog. By statistically analyzing thousands of patients’ electronic medical records, she hopes to discover previously unknown patterns of progression in autoimmune disease. Do lupus patients fall into distinct types that can be identified near the time of diagnosis? Does a change in, say, serum potassium levels predict that a patient with multiple sclerosis will soon suffer a flare-up?

“These are diseases that are very poorly understood,” Saria says. “We want to develop computational frameworks for understanding them in greater levels of detail. Rather than treating the ‘average’ patient, we want doctors to be able to use large-scale population data to tailor decision-making for each individual.”

Saria’s first test case is scleroderma, a rare autoimmune disorder that can cause lung scarring, pulmonary hypertension,and severe thickening of the skin. With the support of a major grant from the National Science Foundation, Saria and her doctoral student Peter Schulam have dived deeply into the electronic records of more than 3,000 patients with scleroderma. Those records have been painstakingly collected over a period of more than 20 years by the Johns Hopkins Scleroderma Center, which is led by a professor of rheumatology at the School of Medicine.

Saria’s team has found several distinct patterns of progression: Some patients, for example, have stable lung function for a long period of time but then abruptly deteriorate; others have lung function that declines in a constant, linear pattern; still others have faulty lung function near the time of diagnosis but then recover.

The next step is to look for biomarkers that might correspond to these subtypes and offer early clues about how a newly diagnosed patient is likely to progress. Those biomarkers might also suggest new targets for drug development. Saria is launching that effort in collaboration with Johns Hopkins rheumatologists Fredrick Wigley, Livia Casciola-Rosen, and Laura Hummers, MS ’10 (BSPH).

“Just by knowing that these patterns exist,” Saria says, “we can now start to ask, ‘Are you the individual who’s going to stabilize, or are you the one who’s going to show active decline?’ If you’re the one who’s going to show active decline, I as a clinician may be more aggressive about starting therapies that might cause more side effects.”