Interpretable convolutional neural networks for sequence-based classification and discovery of plastic-degrading enzymes.
Lee W-H, Dumontet L, Jung K, Lee H, Thapa G
Phytoremediation
Faster discovery of plastic-eating enzymes means the plastic mulch film, nursery pots, and garden bed liners piling up in backyards and landfills have a better shot at being broken down biologically rather than fragmenting into microplastics that end up in your soil.
Some microbes make special proteins that can chew through plastics the way our digestive enzymes break down food. Finding these proteins used to require slow, manual lab work. This AI tool reads the protein's letter-by-letter sequence and predicts which plastic type it can digest—and it can also point to the exact spots on the protein that do the actual breaking, so scientists can trust the result isn't just a guess.
Key Findings
PEPIC classified enzymes across 9 distinct plastic substrate types directly from protein sequences, outperforming state-of-the-art methods with statistically significant improvements in F1-score.
The model was trained on 181 experimentally validated enzymes and ~5,900 homologous sequences, and its highlighted amino acid residues consistently matched known catalytic and substrate-binding sites confirmed by structural modeling.
PEPIC independently identified a previously unannotated enzyme as a potential PET-degrading candidate, demonstrating its value for novel enzyme discovery.
chevron_right Technical Summary
Scientists built an AI tool called PEPIC that can identify enzymes capable of breaking down plastics—just by reading their protein sequences. It outperformed existing methods and even flagged a previously overlooked enzyme as a candidate for degrading PET plastic.
Abstract Preview
The rapid accumulation of plastic waste has emerged as a critical environmental threat, driving the need for scalable and effective biodegradation solutions. Major plastic types, including biopolye...
open_in_new Read full abstractAbstract copyright held by the original publisher.
Was this useful?
Ancient DNA Reveals Pre-Columbian Amazonian Forest Management at Scale
Forests and fruits we romanticize as wild — including many plants now in our kitchens and gardens — may exist in their current abundance precisely because an...