How AI Could Help Solve the World’s Plastic Problem

(Groseclose, T. M.)

How AI Could Help Solve the World’s Plastic Problem

Every year, humanity produces more than 460 million metrics tons of plastic with around 20 million tones of those, ends up as litter in the environment. Where, AI at the sametime considering traditional recycling methods remain far from perfect. Where consuming large amount of energy and often producing downcycled materials, which are often lower in quality than original plastic. As a result, researchers are increasing looking for new ways to address plastic waste more effectively.

A solution that seems promising is the emerging trend from biotechnology, where its artificially engineered enzymes. Scientists hope to use these enzymes to make the close looped plastic possible by breaking these to the original building blocks for them to be reused again and again.

This idea had first gained momentum in 2016, where there was the discovery of the PETase, an enzyme capable breaking down polyethylene terephthalate (PET), one of the world’s most common plastics. That breakthrough showed that biological recycling could be possible.

However, researchers also identified major barriers to using these enzymes at an industrial scale. For enzyme-based plastic recycling to work in the real world, enzymes need several important characteristics, including high catalytic activity, tolerance to different substrates and products, strong thermostability, good expression and solubility, and the ability to function in acidic conditions.

Recent breakthroughs

Since then, advances in AI and protein engineering have accelerated progress in the field.

In 2022, researchers introduced FAST-PETase, an enzyme designed with the help of the machine-learning algorithm MutCompute. FAST-PETase was able to degrade untreated post-consumer PET plastics within a week across a broad temperature range. This marked an important step forward, showing that AI-guided design could create enzymes with practical recycling potential.

Further progress came in 2025, when Wu et al. developed VenusMine, an AI-based tool that searches natural proteins to identify new plastic-degrading enzymes. Using this system, researchers identified an enzyme from the bacterium Kibdelosporangium banguiense, known as Kb PETase. This enzyme showed a melting temperature 32 °C higher than IsPETase and outperformed both FAST-PETase and LCC in degradation tests.

Other studies in 2025 built on this momentum. One research team engineered FAST-PETase for even greater catalytic activity and improved heat tolerance, directly addressing the enzyme’s earlier thermal instability. Researchers have also begun exploring dual-enzyme systems to improve efficiency. For example, combining KL-MHETase with FAST-PETase increased PET depolymerization rates by 2.6 times compared with using FAST-PETase alone.

Another promising strategy involves enzyme immobilization. Kotnis et al. attached FAST-PETase and MHETase to iron oxide nanoparticles, which improved efficiency while also allowing the enzymes to be reused. Reusability is especially important for industrial applications, where cost and scalability are major concerns.

Challenges to scaling up

Despite this progress, several important challenges remain.

One major issue is plastic crystallinity. Even improved enzymes often struggle to break down highly crystalline PET unless the material is pretreated first. This means enzyme-based recycling may still depend on additional processing steps, which could limit efficiency and affordability.

Operating conditions also present a challenge. Many engineered enzymes perform best in tightly controlled environments with specific temperatures, pH levels, and relatively clean plastic feedstocks. Industrial waste streams, however, are much messier. They often contain mixed plastics, contaminant, and other unpredictable conditions that can reduce enzyme performance.

Cost is another significant barrier. Producing enzymes at industrial scale remains expensive, and recent reviews suggest that commercial viability will depend on improving the microbial systems used to express, secrete, and stabilize these enzymes efficiently.

The path forward

To succeed on a larger scale, enzyme-based recycling systems will need to move beyond the lab and into industrial pilot plants capable of handling real-world waste streams. Continued AI-driven discovery platforms like VenusMine could also play a major role by expanding the range of candidate enzymes available for testing and development.

At the same time, dual-enzyme systems and immobilization strategies will likely remain critical. These approaches not only improve degradation efficiency but also make enzymes more practical for repeated industrial use. Finally, lifecycle assessments will be essential to determine whether enzymatic recycling truly offers sustainability advantages over mechanical or chemical recycling methods.

AI-designed enzymes are not yet a complete solution to the global plastic crisis. Still, recent breakthroughs suggest they could become an important part of the answer. If researchers can overcome the remaining technical and economic barriers, AI may help transform plastic recycling from a waste-management challenge into a more circular and sustainable system.

Bibliography:

• Groseclose, T. M., et al. “Recent Advances in Enzyme Engineering for Improved PET Biorecycling.” Communications Materials, 2025.

• Lu, Hongyuan, et al. “Machine Learning-Aided Engineering of Hydrolases for PET Depolymerization.” Nature, vol. 604, 2022, pp. 662–667.

• Organisation for Economic Co-operation and Development. Global Plastics Outlook. OECD, 22 Feb. 2022, oecd.org/en/publications/2022/02/global-plastics-outlook_a653d1c9.html.

• United Nations Environment Programme. “Plastic Pollution.” UNEP, unep.org/plastic-pollution. Accessed 5 Apr. 2026.

• Wu, Banghao, Zhong Bozitao, and Tan Pan. “Harnessing Protein Language Model for Structure-Based Discovery of Highly Efficient and Robust PET Hydrolases.” Nature Communications, 5 July 2025.