The Latest Breakthrough In Understanding Diabetes Was Made By An Algorithm

Researchers now believe there are three different kinds of type 2 diabetes–a result discovered with help from machines combing through reams of medical data.

The Latest Breakthrough In Understanding Diabetes Was Made By An Algorithm
[Top Photo: munalin via Shutterstock]

With the cost to sequence a human genome dropping by the day and medical records finally going digital, public health experts are excited for a new era of personalized, or “precision,” medicine–a big data future in which there is no “average” patient, only individual patients with unique genes, environments, and lifestyles. As a measure of this excitement, this year, President Obama launched a $215 million initiative that will create a health database from 1 million volunteers that is unprecedented in detail. Breakthroughs in prevention, understanding, and treatment of disease are hoped.


Though there’s both hype and a lot of genuine promise, the field of precision medicine is still in its nascency. Genetic sequencing has helped in the diagnosis and treatment of rare genetic disease and is beginning to be important in the treatment of some cancers, such as lung cancer or brain tumors.

Now a recent study, published in the journal Science Translational Medicine, demonstrates the broader promise of precision medicine beyond genome sequencing–and in understanding an extremely common disease: type 2 diabetes.

Almost 1 in 10 Americans have type 2 diabetes, and many more are at risk. Yet it’s a poorly understood disease: Its causes, symptoms, and complications are diverse and hard for doctors to predict. By mining a database of clinical and genetic data from more than 2,500 diabetes patients, researchers Icahn School of Medicine at Mount Sinai Medical Center have now actually identified some patterns that an entire field of doctors have not: They found there are actually three distinct sub-types of type 2, each of which have very different health implications.

“This is the first tangible demonstration of precision medicine that could be applied to a more common, complex disease,” says study author Joel Dudley, director of biomedical informatics at Mount Sinai (and one of Fast Company’s Most Creative People of 2014), a major hospital in upper Manhattan.

They were able to do this with access to a still relatively rare collection of thousands of Mount Sinai patients who volunteered to give their health charts and genetic data to the hospital for researcher efforts (See “In The Hospital Of The Future, Big Data Is Your Doctor“). Usually, doctors just look at a few blood tests–such as blood sugar and insulin levels–when monitoring diabetes patients. Instead, the researchers used computer modeling to map how similar each patient was to each other (a “patient-patient similarity network”), based on every piece of health data–height, weight, blood platelet counts, and hundreds of data points that human doctors alone could never process. The result was the map, seen above, that shows Mount Sinai patients map into three distinctive clusters, or “sub-types.”

The bigger question, of course, is: Do these sub-types matter for a patient’s health? The study found that they probably do. Patients in one subtype were more likely to suffer from cancer and cardiovascular disease; in another subtype, they were at higher risk for kidney disease and eye complications; finally, in the third sub-type, allergies, neurological diseases and HIV infections were bigger problems. Even more importantly, when the researchers mapped patients’ genomes onto the network–they found unique gene variants associated with each sub-group which helped explain some of the differences between them.


“The fact that these genetic factors matched up so nicely with the clinical factors suggests that there’s actual biology underlying the differences between these patients,” says Dudley, as opposed to other explanations (like lifestyle choices.)

Diabetes experts weren’t too surprised by the patterns in type 2 diabetes, according to Dudley–they already knew them in their gut and had seen patients of each “sub-type.” “But they didn’t have the ability to visualize it in this way,” Dudley says.

The next step will be to replicate the study in other patient populations beyond upper Manhattan’s unique mix of wealth Upper East Side residents and lower-income minority residents of East Harlem. A larger, long-term goal is to do a more proactive study–to see if, based on their genes, the unique risks faced by newly-diagnosed diabetes patients can be predicted and ultimately if their treatment plans can be customized.

Ultimately, Dudley envisions transferring these same methods to a host of other diseases to make discoveries that weren’t possible unless you’re looking at millions of data points at once. He envisions a “google maps” for a disease like diabetes, where health tests would show where any given patient was on a broad range of possibilities for how the disease would progress. But this will be slow–for now, not many kinds of data sets like the Mount Sinai one exist. The White House initiative should add at one important one.

About the author

Jessica Leber is a staff editor and writer for Fast Company's Co.Exist. Previously, she was a business reporter for MIT’s Technology Review and an environmental reporter at ClimateWire.