11-year-old Bertrand (Buddy) Might was ill. He was down with a mysterious condition which made him unenergetic and confused. His fluid-filled lungs prevented his breathing completely when he was admitted at the Children’s Hospital of Alabama this May. Even though the doctors managed to insert a tube to drain his lungs, Buddy was left with a swelling stomach and yellow substance could be found in his bones. There was no diagnosis. 

Like most children who believe their parents have super-power; Buddy also has an extraordinary dad. His father, Matt Might is an artificial intelligence (AI) expert who heads the Hugh Kaul Precision Medicine Institute at the University of Alabama (UAB). Matt’s work includes building systems to automate the cleaning and analyses of biomedical data that are streaming into the institute. 

Sitting by Buddy’s bedside, Matt decided to have a try with AI and see if the software could shed light on Buddy’s unexplainable symptoms. Known as mediKanren, the algorithm was developed as part of a project funded by the National Institutes of Health, to ambitiously pool all the biomedical data, ranging from proteins, genes, drugs, to disease symptoms, patient outcomes and so on, gathered by universities and research labs over the years. Matt believes physicians and researchers will be able to benefit mediKanren, because previously, these data were scattered in more than 100 independent databases. 

The pioneer patient 

Buddy became mediKanren’s first patient. He is also the first patient in the World with a double mutation in the NGLY1 gene. NGLY1 is responsible for producing an enzyme that assists in the recycling of cellular waste. Buddy’s deprivation of this particular enzyme had resulted in an untreatable neurodegenerative disease that left him wheelchair-bound and a series of symptoms like developmental delays and incapable to create tears. 

As Matt entered Buddy’s symptoms into mediKanren, it collated them together to produce some of the most viable explanations. With his son’s mutation at the back of his mind, Matt knows Buddy’s latest illness has nothing to do with cancer or autoimmunity. Matt’s AI career is pretty much knitted to Buddy’s medical condition ever since his mutation was confirmed in the year 2010. 

Matt switched his focus from cybersecurity to healthcare fundamentally because he knows the rarity of Buddy’s condition would make him an outliner, shall he be in the hands of a public health system. Matt began by partnering the pharmacy school at the University of Utah where he teaches, to develop an algorithm that repurposed existing drugs to cater to the needs of patients with genetic mutations like Buddy. Matt’s effort soon attracted attention from other academia and he rose to fame when former US President Barack Obama invited him to be part of a precision medicine initiative, which Matt undertook until 2018.

Benefitting other patients 

It turned out that Buddy was suffering from a septic shock this May. Nevertheless, his condition did not improve even though he had his lungs drained and two surgeries to remove the yellow substance from his thigh bones. Buddy’s symptoms were coded by mediKanren each time when he was admitted into the emergency ward or seen by different groups of specialists. Eventually, mediKanren led Matt to a list of possible microbes, including a kind of bacteria called Pseudomonas. 

Indeed, when the lab results were back, Buddy’s lung fluid shown to have a heightened level of Pseudomonas and E.coli. MediKanren’s prediction had thus guided Buddy’s doctors to prescribe meropenem, an antibiotic that is capable to wipe out both Pseudomonas and E.coli. Buddy’s condition, was finally, stabilized. 

The beauty of mediKanren lies in its power to provide crucial information and highlight the sources of the information to medical professionals so that they can follow its logic. As John Bowers, an AI researchers at the Berkman Klein Center for Internet and Society at Harvard University told Stat, “part of what’s so exciting about Matt Might’s system is that it seems to have a deep respect for causality and the role of the human analyst. It shows how a human analyst can benefit from the giant troves of associational predictions and data, but it doesn’t subjugate the analyst to the machinations of the AI system”. 

70 other people were diagnosed with double NGLY1 mutation after Buddy and mediKanren continued to help patients who approached UAB because of their unexplainable symptoms. The team believes while mediKanren is showing potential, there is still room for improvement, especially when dealing with patients whose conditions are deteriorating within a short period of time. 

Author Bio
synthetic gene empathy chinese artificial intelligence data medicine healthcare ai

Hazel Tang

A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.