Researchers Investigate the Effects of Medication Interaction when Treating Kidney Disease

Researchers at Stony Brook University are investigating how drug interactions could affect a patient’s risk of kidney failure. 

“The initial problem is that if a patient has AKI, there may be a high risk for the patient to have kidney failure, like in two years or five years,” said Tengfei Ma, assistant professor in the Department of Biomedical Informatics in Stony Brook’s Renaissance School of Medicine and College of Engineering and Applied Sciences, and lead author of the project’s upcoming study. “But then we found that if a patient has AKI, maybe the patient needs to take a lot of different medications. Some of the medications, when they are combined together, can have some adverse reactions.” 

AKI stands for Acute Kidney Injury, and it refers to a period of sudden kidney failure. After a patient has AKI, the risk increases for them to be diagnosed with Acute or Chronic Kidney Disease. 

The specific medication interactions the researchers are looking at are called “Drug-Drug Interactions,” which involve any effect caused by a reaction between two or more medications in the body. 

The researchers use data from Electronic Health Records (EHR), as well as knowledge graphs, which track information through a network of nodes – or points that represent individual entities. 

“So you have some nodes as maybe drugs, proteins, genes, and then you have the relationship between the nodes, the drug-drug interaction,” Ma said. “You can capture the relationships among all kinds of nodes, like which drug can impact this gene.” 

Ma and his research team combine the data from the EHRs with knowledge graphs to analyze whether possible interactions between the drugs a patient is taking could affect the patient’s diagnosis. 

“We want to extract these signals from the patient’s EHRs,” Ma said. “But one issue of this kind of information tracking is that the EHR data is kind of messy. So we want to combine the signals from both the EHRs and also some knowledge graphs. We already have some knowledge graphs about drug-drug interactions, so you can actually utilize the existing drug-drug interaction to do the prediction of unknown drug pairs.” 

Drugs are prescribed to prevent adverse health effects, but when multiple drugs are prescribed and taken at once, interactions between them could cause unintended outcomes. 

“We want to predict kidney failure risk,” Ma said. “So if you don’t take any medications, this risk will not be changed. But if you take one medication, this risk might decrease. But if you take the two drug combinations together, the risk will probably increase again. So that’s also one kind of drug-drug interaction. If you take one drug, generally it’s very useful, but if you take another drug, this drug may increase or decrease the effect of another drug.” 

Seed funding has been granted for this project, “Integrating EHRs and Knowledge Graphs to Detect Drug-Drug Interactions Affecting Post-AKI Kidney Outcomes,” with some preliminary results already published.

“The data set we used was quite small, so it’s more like a prototype,” Ma said. “I think the outcomes just show that using both EHRs and knowledge graphs can generally give you better performance than using isolated HRC and knowledge graphs.” 

Ma and his team are now working to expand their research to cover a larger data set. 

“Currently, we are extracting some larger genetic data sets. We want to start from the genetic effects of the single medication, so that we can build some foundations for future analysis of drug pairs. And of course, if you have a larger data set, you also need to deal with a larger knowledge graph. So we need to redo the previous experiments combining the knowledge graph and the EHRs again, to get the new results.” 

Ma hopes the outcome of their research can lead to changes in how the medical field treats patients. 

“If we succeed, our results may change some guidelines for the clinicians when they deal with AKI patients,” he said. “And this methodology is actually not only applied to AKI patients, but it can also be applied to any kind of disease. So when clinicians think about some medication combinations, they will think, ‘Okay, what kind of risk do we have if I give the prescriptions together?’ That will also help with the safety of the patients.” 

— Mattia Stradella