From prediction to validation: Critical assessment of computational plastizyme discovery pipelines for plastic biodegradation

Plastic biodegradation research has rapidly shifted from culture-based enzyme isolation to large-scale computational mining of environmental omics datasets. However, heavy reliance on sequence homology has inflated predictions of plastic-degrading capacity, as conserved catalytic motifs and structural folds rarely guarantee substrate specificity toward synthetic polymers. This review provides a critical and integrative analysis of computational Plastic Degrading Enzyme (PDE) discovery, linking environmental omics, bioinformatics annotation, structural prediction, molecular docking, machine-learning prioritization, and experimental validation. Unlike previous reviews that treat these components independently, we evaluate them as interconnected layers within a function-centred discovery framework. We highlight key limitations including assembly bias, database misannotations cascades, simplified polymer modelling in docking, and uncertainty in AI-driven predictions and propose best practices to reduce false-positive inflation. By integrating structural plausibility, catalytic geometry, ecological context, and biochemical verification, this review reframes PDE discovery from homology-driven inference to multi-layer functional validation, offering a more reliable roadmap for engineering sustainable biocatalysts for plastic waste mitigation.