Metagenomic exploration and computational prediction of novel enzymes for polyethylene terephthalate degradation

As a global environmental challenge, plastic pollution raises serious ecological and health concerns owing to the excessive accumulation of plastic waste, which disrupts ecosystems, harms wildlife, and threatens human health. Polyethylene terephthalate (PET), one of the most commonly used plastics, has contributed significantly to this growing crisis. This study offers a solution for plastic pollution by identifying novel PET-degrading enzymes. Using a combined approach of computational analysis and metagenomic workflow, we identified a diverse array of genes and enzymes linked to plastic degradation. Our study identified 1305,282 unmapped genes, 36,000 CAZymes, and 317 plastizymes in the soil samples were heavily contaminated with plastic. We extended our approach by training machine learning models to discover candidate PET-degrading enzymes. To overcome the scarcity of known PET-degrading enzymes, we used a Generative Adversarial Network (GAN) model for dataset augmentation and a pretrained deep Evolutionary Scale Language Model (ESM) to generate sequence embeddings for classification. Finally, 21 novel PET-degrading enzymes were identified. These enzymes were further validated through active site analysis, amino acid composition analysis, and 3D structure comparison. Additionally, we isolated bacterial strains from contaminated soils and extracted plastizymes to demonstrate their potential for environmental remediation. This study highlights the importance of biotechnological solutions for plastic pollution, emphasizing scalable, cost-effective processes and the integration of computational and metagenomic methods.