"Towards Quantitative Structure-Property Relationships for Polymer Biodegradability"
Abstract: The exponential rise in the production and use of plastics, particularly in single use applications, has led to a dramatic increase in their environmental prevalence and problems with plastic waste management. One necessary component of the solution to this challenge is developing plastics that degrade more effectively when they are accidentally released in to the environment, an unavoidable occurrence at some level in any practical waste handling system. An important part of the solution to this problem is the engineering of novel polymers with controlled biodegradation in the natural environment. Biodegradation is a property of further interest for compostable systems and systems that may be discharged into municipal water treatment facilities.
Although biodegradation is believed to be a function of chemical structure and therefore should be amenable to quantitative structure-property methods such as group contribution theory or more recent machine learning approaches, the field is plagued by a lack of data. Herein, we report the adaptation of the clear zone assay from molecular biology to the high-throughput screening of biodegradation that can overcome long test times of standardized methods and enable a large biodegradation data set to explore structure-property relationships. We report the synthesis and biodegradation testing of over 1,000 different polyesters. Library design incorporates a wide variety of different chemical functionalities to specifically probe as diverse a chemical space as possible within the class of polyester chemistries. The study of the large library enables us to analyse the impact of a variety of different functional groups on polymer degradation across a large number of polymers, extracting a number of useful trends in chemical structure. This large data set is then analysed using simple regression and random forest classifier (RFC) machine learning methods to attempt to classify polymers as biodegradable or not biodegradable. We explore the relative role of chemical structure as well as property descriptors (i.e. crystallinity) and molar mass in the efficacy of predictions, demonstrating accuracies above 82% for many approaches.
Initial work focused on polyesters has now been expanded to polyurethanes, and extension from biopolymers to terpolymers provides an added test of the ability of the model to extrapolate to more complex chemical structures. As a part of these studies, we have also revealed that certain commonly-used catalysts may be toxic to cells, having a large impact on microbial growth. Finally, we are in the process of extending our assay to a wide variety of microorganisms, aiming to provide “biodegradation fingerprints” that can be used to predict environmental fate in a wide variety of specific environments.
Bio: Bradley Olsen is the Alexander and I. Michael (1960) Kasser Professor in the Department of Chemical Engineering at MIT. He earned his B.S. in Chemical Engineering at MIT, his Ph.D. in Chemical Engineering at the University of California – Berkeley, and was a postdoctoral scholar at the California Institute of Technology. He started as a professor at MIT in December 2009. Olsen’s research expertise is in materials chemistry and polymer physics, with focused activities in the areas of molecular self-assembly, polymer networks, natural and sustainable materials, and polymer informatics.
