New Animal-Free Model May Help Predict How "Forever Chemicals” Behave in the Human Body

Per- and polyfluoroalkyl substances, better known as PFAS, or "forever chemicals” can linger in the human body for years, yet they clear from a rat's system in days or weeks. This enormous species gap makes it difficult to use animal data to predict PFAS health risks in people.

Researchers at the U.S. Environmental Protection Agency tackled this challenge by developing a new computational model called HTTK-PFAS, which combines human cell-based laboratory measurements with a machine learning model that predicts how long PFAS persist in different species. The machine learning predictions are used to model whether a given PFAS is likely being pumped back into the bloodstream or flushed out. This new approach is a dramatic improvement over the standard animal method and is now freely available to other researchers.

This work supports a critical shift away from animal testing. Because PFAS behave so differently across species, animal studies have limits for predicting human risk. Tools that predict human outcomes directly using human cell data and artificial intelligence reduce the need for animal tests while enabling faster safety assessments across the tens of thousands of "forever chemicals” that exist. By combining machine learning with human cell-based data, researchers built a tool that better predicts how PFAS behave in the human body to advance chemical safety science while reducing the need for animal testing. The Physicians Committee hosts webinars and training courses in nonanimal regulatory test methods like this through our NAM Use for Regulatory Application program.