An AI-driven closed-loop framework: A novel enzymatic biochar enabling near-complete degradation of biodegradable plastics.

The anaerobic co-digestion (AcoD) of food waste (FW) and biodegradable plastics (BPs) is a promising waste-to-energy strategy, yet it remains severely bottlenecked by asynchronous hydrolysis. Overcoming this limitation via traditional trial-and-error optimization is prohibitively slow and expensive. To address this, an integrated machine learning (ML) framework was developed, coupling predictive modeling with rigorous experimental and mechanistic validation. Virtual screening initially identified mesophilic enzyme-loaded biochar (BC) as the optimal intervention. Consequently, a novel proteinase K-loaded BC (PKBC) was synthesized, which successfully achieved up to 90% degradation of 2-mm BPs films in a 120-day trial. A high-precision random forest model further mapped the system dynamics, confirming that degradation is positively driven by reaction time and methane yield, but tightly constrained by larger particle sizes. Providing a fundamental basis for these predictions, multi-omics analyses revealed that PKBC triggers a targeted metabolic cascade. This cascade reprograms the microbial community to accelerate lactate-driven BPs cleavage and maximize hydrogenotrophic methanogenesis. Ultimately, this work delivers a highly efficient solution for mixed waste treatment and establishes a transferable AI-first paradigm for the intelligent design of complex bioprocesses.