The Scripps Research Translational Institute and Nvidia are teaming up to apply artificial intelligence to genomic and health sensor data, the organizations announced Tuesday.
The goal of the new partnership is to detect and prevent diseases by applying machine learning and deep learning to data from genome sequencing and digital health sensors. Because sensor and genomic data are so plentiful, the only way to truly understand and analyze them is with machines, said Dr. Eric Topol, founder and director of the Scripps Research Translational Institute.
"It's a formidable task that no one has yet achieved," he said. The amount of genomic data doubles every seven months, according to the organizations. By applying deep learning to that data, Scripps and Nvidia researchers will be able to better detect mutations and also lower the cost of genome sequencing.
At first, the organizations will focus on heart rhythm disorders, drawing data from both genetic sequencing and sensors, including the sensors on wearables like Fitbits and Apple Watches. Apple recently announced its newest Apple Watch will be able to conduct electrocardiograms.
Scripps is already drawing on Fitbit data for part of the National Institutes of Health's All of Us research program, intended to advance precision medicine. The program, with $1.45 billion in funding over 10 years, involves collecting the health records and DNA of 1 million people. Scripps is looking into how the researchers can incorporate into the initiative Fitbit measurements of biological processes, such as sleep and heart rate.
Through the Nvidia partnership, researchers might be able to identify the electrocardiographic markers for what happens right before atrial fibrillation, Topol said. That way, the condition could be prevented before it even begins.
Apple and Stanford are already studying something similar with the Apple Heart Study. Researchers are collecting data to figure out how heart rate sensors can play a role in precision health.
Harnessing artificial intelligence and data could cut healthcare costs.
"When big data and modern technology are leveraged to predict health outcomes, the resulting information can be used to improve health across the population," said Mark Nathan, founder and CEO of health insurance software company Zipari.
Scripps and Nvidia hope to move on to other conditions, such as high blood pressure and diabetes, where continuous monitoring could change treatment.
Researchers will also apply deep learning and other AI-based techniques to existing datasets, such as the Scripps' whole-genome sequences of healthy elderly people.
Much of the work involving AI in healthcare so far has been related to imaging. But for personalized medicine to truly take off, healthcare organizations will need to apply the technology to other datasets, said Kimberly Powell, Nvidia's vice president of healthcare.
Sensors will be key to this work, and Powell compared their use in healthcare to their use in driving and cars. Just as self-driving cars rely on sensors to create a safer driving environment, healthcare can rely on sensors to create a safer environment too.
That information becomes useful when machine learning algorithms and others extract meaning, detecting what humans can't see.
"The problem with humans—experts even—is you can only see certain things because you just don't have the capability that really good machine algorithms have," Topol said.
