Chris Bizon’s AI work at RENCI solves research puzzles
Tools powered by artificial intelligence help scientists discover information vital for daily life buried in vast collections of data.
Chris Bizon spends his days assembling puzzles with millions of pieces.
Bizon ’92 directs a team at Carolina’s Renaissance Computing Institute that creates ways to comb through millions of research papers, laboratory results, government records and biomedical data scattered across countless databases.
What the team finds is greatly relevant to daily life. New uses for existing medicines. Deadly hazards along North Carolina roads. Hidden environmental contamination. Useful scientific information from decades of handwritten field notes.
“No human could complete such puzzles alone,” said Bizon, who directs RENCI’s analytics and data science group. “The tools have changed over time, from numerical simulation to machine learning to AI, but the thrust of using computers to make research work better is the same.”
The team collaborates with researchers across fields such as medicine, transportation, art history and environmental science. Their common goal is answering difficult questions by applying computational methods, including machine learning, artificial intelligence and knowledge graphs. A knowledge graph is a structured way of organizing information by connecting people, places, diseases, genes, drugs and other entities through their relationships.
Improving human health
Bizon’s team has contributed to the National Center for Advancing Translational Sciences Biomedical Data Translator project, which integrates large collections of biomedical information so scientists can convert discoveries into treatments, tools and improved health.
The group also co-developed MATRIX, the Machine learning/AI-enabled Therapeutic Repurposing In eXtended uses platform. Every Cure, a $100 million initiative, uses the MATRIX platform to seek new uses for existing drugs to treat diseases without effective therapies. By uncovering hidden links among genes, diseases and treatments, MATRIX helps researchers identify promising treatments that might otherwise remain buried in scientific literature.
Related work includes an open-source knowledge graph known as ROBOKOP or Reasoning Over biomedical objects linked in Knowledge Oriented Pathways. ROBOKOP was developed with collaborators including Alexander Tropsha, K.H. Lee Distinguished Professor at the UNC Eshelman School of Pharmacy. Researchers recently used ROBOKOP to identify existing drugs that may help treat rare diseases.
Enhancing driver safety
A state-funded effort with the North Carolina Department of Transportation and the UNC Highway Safety Research Center used machine learning to identify roadside hazards across rural North Carolina. Survey vehicles videotape the edge of the state’s rural roads. RENCI’s system analyzes those images and flags objects that could become deadly obstacles if a vehicle leaves the roadway.
“Most of the fatalities in North Carolina are single-car departures from the road hitting an object,” Bizon said. Creating a statewide catalog of hazards helps officials prioritize safety improvements.
Understanding nature
Another project is a collaboration with the North Carolina Museum of Natural Sciences. Researchers there needed help transforming handwritten field notes about fish species into searchable, usable data. Bizon’s team is using vision models and AI agents to extract information and georeference historical records.
“If we just woke up one day, I doubt we’d have said, ‘We should build a tool to do this,’” Bizon said. “But when somebody says, ‘I need a way to do something better and faster,’ it becomes something that we get excited about.”
Bizon grew up in Elon, North Carolina. After attending Carolina, he earned a doctorate in physics from the University of Texas at Austin. There, he developed computer simulations to model physical experiments involving fluid flows and other complex systems. Later, at GlaxoSmithKline, he used machine learning to predict which potential drug compounds would likely succeed in laboratory testing.
AI is changing how his team develops software. Increasingly, programmers use AI systems to generate code, accelerating development while still requiring human oversight.
“Our work on these tools is based on the researcher’s needs,” Bizon said. “We have to express what we find in a way that is useful.”
