Insights into selective volatile fatty acid production from pretreated food waste unveiled through kinetic modeling and machine learning.

The industrial potential of food waste (FW) acidogenesis is constrained by the difficulty in separating mixed volatile fatty acids (VFAs). Selective production offers a solution, yet mechanistic knowledge gaps impede precise control over product selectivity. To address this, pretreatment conditions were systematically investigated, with the aid of kinetic modeling and machine learning to identify key factors governing product selectivity. A regulatory framework based on pH and thermal hydrolysis pretreatment (THP) was established for selective VFA production from FW-based fermentation. Highest selectivities of 96.0 %, 52.1 %, and 92.6 % (of total VFAs) were achieved for acetate, propionate, and butyrate, respectively, with corresponding yields of 184.0, 184.5, and 338.2 mg COD/g VS. Kinetic modeling further indicates that the theoretical selectivities for propionate and butyrate can reach 56.9 % and 96.3 %, respectively. Further analysis revealed that the pH and substrate composition reshaped by pretreatment contributed to a dual-control mechanism: pH acted as an activity switch, determining pathway feasibility and activation order, while substrate composition functioned as a metabolic-intensity modulator, allocating flux among active pathways. This proposed mechanism was then quantified by kinetic modeling and machine learning, which clarified when and how selective production was achieved. Although this approach involves a yield-selectivity trade-off, it establishes a reliable benchmark for future work aimed at concurrently improving both, representing a decisive step toward industrial deployment.