Murdoch University Researchers Use Machine Learning to Find Plastic-Degrading Enzymes

Machine learning is filtering through millions of enzyme structures stored in global databases to identify biological candidates capable of breaking down plastics and per- and polyfluoroalkyl substances (PFAS), nanowerk.com reported October 4, 2026. Scientists at Murdoch University’s Bioplastics Innovation Hub are combining biochemistry with computational tools to bypass the slow, manual search for effective bioremediation agents. The key to finding the right biological fit to combat each pollutant was likely hidden within existing data, according to reviews published in Nature Reviews Earth & Environment.

Targeting Persistent Pollutants With Evolutionary Solutions

Joseph Boctor, a PhD candidate at Murdoch University, argues that attempting to overengineer synthetic enzymes ignores natural evolutionary processes. I strongly advocate that overengineering enzymes is a bad starting point that overlooks millions of years of evolution that have already produced lots of potential solutions to these contaminants, Mr Boctor stated in his published review. The computational pipeline developed by the research team uses data from previously characterized enzymes to predict how natural structures interact with target pollutants. These tools aim to address contaminants currently active in the ecosystem rather than solely focusing on future bioplastic alternatives.

Microplastics and PFAS Moving From Agriculture to the Human Food Supply

Persistent pollutants present severe biological risks because they mimic human hormones and disrupt normal health functions. PFAS, microplastics and other persistent pollutants are not only industrially favourable but biologically active, Mr Boctor said. A comprehensive review conducted last year revealed that agricultural soils contain approximately 23 times more microplastics than oceans. Researchers have detected microplastics and nanoplastics inside edible crops including lettuce, wheat, and carrots. Soil samples also contained phthalates, which are linked to reproductive issues, alongside polybrominated diphenyl ethers (PBDEs), which are neurotoxic flame retardants associated with heart attacks, strokes, and early death.

Testing and Scaling Enzymes for Real-World Bioremediation

While machine learning successfully narrows millions of biological candidates down to viable matches, physical laboratory validation remains essential. The Bioplastics Innovation Hub team is currently focusing its resources on testing, validating, and scaling the computationally identified enzymes. We need to look in the right place, and leveraging machine learning tools we are able to mine through millions of unexplored biological data to find the right candidate for the relevant task, Mr Boctor noted regarding the shift toward data-driven environmental cleanup.