Researchers Apply Data Expertise to Advance Knowledge of Kidney Disease
The Centers for Disease Control and Prevention estimated in 2023 that 14% of Americans—more than 35 million people—suffer from chronic kidney disease. Researchers at University of Tennessee, Knoxville, and the University of Tennessee Health Science Center in Memphis are using analytics and artificial intelligence to advance understanding of the disease and explore options for potential treatment.
Michael A. Langston, a professor of computer science in UT's Tickle College of Engineering, is working with Csaba P. Kovesdy, Fred Hatch Professor of Medicine in Nephrology at UTHSC, to study associations between risk factors, exposures, and clinical outcomes in patients with various stages of chronic kidney disease. In a separate project, Langston is partnering with Keiichi Sumida, a professor of nephrology at UTHSC, to explore connections between bacteria and fungi in the bloodstream and inflammation in end-stage renal disease.
Langston applied his expertise in data science, graph algorithms, machine learning, and statistical software for both research collaborations.
"For more than 25 years, the Veterans Administration has compiled data from daily clinical operations," Kovesdy said. "It's a huge amount of information, from 3.5 million patients, and we needed a colleague to use novel graph theoretical methods to explore it. We approached the project with the desire to see if the data it would lead us to previously unknown conclusions."
Data leads to new insights
Langston and his team began searching for commonalities in the data after all personal identifiers had been removed to protect individual patient privacy. They discovered that across all patients congestive heart failure, heart attacks, peripheral vascular disease, and cardiovascular disease formed a dense subgraph that indicated highly interconnected disease relationships.
"When it comes to an affliction like kidney disease, age should be a telling factor," Langston said. "After controlling for diet, pollutants, and so forth, the older you get, the more likely you are to have kidney problems. However, if you develop chronic kidney disease for some reason, whether you're young or old is no longer very useful in forecasting your longevity. Instead, we find that a constellation of cardiac factors is a much better predictor."
Data analysis was also crucial to Sumida's work researching bacteria and fungi in the bloodstream of patients with end-stage renal disease. For this project, Langston and his team analyzed data from 1,000 dialysis patients nationwide—information drawn from frozen samples of their blood as well as their clinical care and clinical outcomes. The team worked to identify bacterial and fungal DNA signatures associated with a higher risk of cardiovascular death.
Together they identified specific patient characteristics, such as a low count of the white blood cells called neutrophils, associated with fungal DNA in the blood, Sumida said. These fungal communities are associated with cardiovascular mortality.
While the projects are observational studies and not laboratory research, studying this type of data helps researchers identify associations, hypothesize about plausible explanations, and lay the groundwork for other investigations. Future work could include experimental studies or clinical trial-like interventions where researchers try to lower the amount of these types of fungi in patients' blood and monitor potential improvement in their health outcomes.
"Think of a sea of variables, literally thousands of them, and we're looking for latent relationships hidden deep within the data," Langston said. "We're letting the data speak to us, generating compelling hypotheses not limited by preconceived ideas."
Provided by University of Tennessee at Knoxville
