A Computational Prediction of Novel PET-Degrading Enzymes: Pathway to Sustainable Plastic Management

Polyethylene terephthalate (PET) is a commonly employed polyester owing to its favorablecharacteristics and affordability. Nevertheless, PET plays a substantial role in escalating plastic wastepollution [۱],[۲]. Despite its crucial role, this has severe environmental consequences due to the slowPET degradation rates. The rise of PET poses a growing environmental threat, endangering diverse lifeforms including humans. Existing methods for managing plastic waste are notably inefficient,emphasizing the necessity for creative solutions, such as PET biodegradation. Enzymes capable ofdegrading plastic, known as plastizymes [۳], primarily target high-molecular-weight polymers, such aspolyethylene terephthalate (PET). However, isolating these PET-degrading enzymes remains achallenge, primarily due to the cultivation difficulties associated with many microorganisms. This studyembarks upon this environmental challenge. Our research employed an innovative computationalstrategy to identify novel PET-degrading enzymes that break down PET. Given the limited dataset ofknown PET-degrading enzymes, we employed a Generative Adversarial Network model to augment thedataset using synthetic but realistic protein sequences. For an in-depth understanding of these enzymes,we extracted feature embeddings from the evolutionary scale model to facilitate the training of multipleclassifiers, including the Support Vector Machine, K-Nearest Neighbors, and Random Forest models.These results have led to the identification of novel PET-degrading enzymes. Their validation wasfurther supported by the analysis of the active sites, crucial amino acid compositions, and ۳D structurecomparison. Our study establishes a stage for substantial advancements in the plastic degradationindustry by employing computational methodologies for plastizyme prediction, and emphasizes thepotential of metagenomic approaches in environmental remediation research.

کلیدواژه ها:

metagenomics ، machine learning ، PET-degrading enzymes ، computational prediction