Machine learning predicts biodegradation rates of PHBV bioplastics in real-time

On July 7, 2026, researchers from the Agricultural University of Athens unveiled a groundbreaking machine learning tool designed to predict how quickly biodegradable plastics break down in nature. This innovative study, published in Polymers, focuses on PHBV (poly(3-hydroxybutyrate-co-3-hydroxyvalerate)), a biopolymer recognized for its potential as a sustainable alternative to traditional fossil-based plastics.

Advancements in Biodegradable Plastics

The traditional method for assessing the biodegradation of plastics often requires extensive lab work that can take months or even years. The new machine learning model offers a faster approach, yielding nearly instant predictions for PHBV biodegradation outcomes. This is particularly significant in humanitarian contexts where waste management systems are often lacking.

The research team, led by Chrysanthos Maraveas, compiled a comprehensive database from 13 peer-reviewed studies conducted over three decades. This extensive dataset includes 93 experimental cases and over 1,300 individual biodegradation measurements based on CO₂ evolution.

Machine Learning Techniques and Findings

Two machine learning techniques, Random Forest and XGBoost, were employed to analyze the data. Both models demonstrated impressive predictive accuracy, achieving R² values between 0.95 and 0.97, indicating their reliability in forecasting biodegradation rates on unseen data.

Key factors influencing biodegradation time were identified, with biodegradation duration being the primary predictor. Additional factors included temperature, the ratio of the polymer's components, the degradation mechanism, microbial community type, and the content of additives. These findings underscore the complex interplay between material design and environmental conditions.

Public Access to Predictive Tools

The Random Forest model has been made publicly accessible as a free, interactive web tool on the Jaqpot platform. This tool allows researchers and manufacturers to input specific formulation and environmental parameters, providing quick predictions on biodegradation rates. This initiative supports a “safe-and-sustainable-by-design” philosophy for developing next-generation biodegradable materials.

For more information, refer to the study by Marianna I. Kotzabasaki et al, titled Machine Learning Methods for Mineralization-Based Biodegradation Prediction in Polyhydroxyalkanoate-Based Biopolymers: Insights from Lab-Scale Experiments, published in Polymers in 2026.