Abstract
Microbial electrorespiration harnesses bacteria to drive reductive dechlorination, offering a sustainable method to remediate environments contaminated with persistent chlorinated organic pollutants (COPs). However, aquifers' complex hydrogeological and hydrochemical conditions, combined with uneven COP distribution, create substantial spatial and temporal variability in biochemical reactions, environmental factors, and microbial communities. Traditional trial-and-error experiments are labor-intensive and slow, impeding the quick identification of conditions that accelerate dechlorination rates. Here we show that a machine learning framework, integrating experimental design with cathodic biofilm data, uncovers key interrelationships among environmental variables, dechlorination kinetics, electrochemical properties, and functional microbes, enabling rapid optimization of bioelectrodechlorination. Trained on literature-derived datasets using models such as extreme gradient boosting, random forest, and multilayer perceptron, this framework identifies temperature and cathode potential as primary drivers in experimental design while highlighting key biofilm genera, including Clostridium, Desulfovibrio, Dehalococcoides, Pseudomonas, Dehalobacter, Arcobacter, Lactococcus, and Geobacter. It supports inverse design to determine optimal parameters-such as cathode potential, temperature, and additives-for dechlorinating representative COPs, including tetrachloroethene, trichloroethene, and 1,2-dichloroethane, achieving reaction rate predictions with errors below 6 %. This approach surpasses conventional methods by increasing efficiency, cutting costs, and accelerating bioremediation without extensive laboratory testing. By incorporating microbial community insights into predictive models, our data-driven strategy advances the scalable application of microbial electrorespiration for COP-contaminated water remediation and paves the way for broader bioelectrochemical uses in environmental engineering.