A comprehensive guideline for hybrid modeling of engineered microbial processes

Microbial processes have been extensively engineered to remove contaminants and recover value-added products. Despite their practical significance, these processes present unique challenges in both design and operation due to the inherent variability and complexity of microbial populations and communities. As the driving force of engineered microbial systems, the activity of microbial populations and the structure of their communities remain difficult to control and model. Hybrid models combine the interpretability of mechanistic models with the flexibility of data-driven models, offering the potential for more robust predictions. However, their application in engineered microbial processes remains underdeveloped. To assess the current state of hybrid modeling in relevant studies, we conducted an extensive literature review that identified only 52 qualified articles over the past 30 years, including 32 hybrid models reported in 30 articles published within the last five years. We systematically examined key stages of hybrid modeling-data collection, data processing, and model construction-and revealed critical challenges that were frequently overlooked. Among these, hyperparameter tuning was missing in 21 studies, followed by data leakage, with 16 studies lacking explicit statements and 4 demonstrating potential risks in our analysis, both of which significantly compromise model performance and reliability. To better provide technical support and highlight common pitfalls, this review summarizes these issues and proposes a standardized protocol tailored to engineered microbial processes. It further delivered detailed analyses and practical recommendations for essential modeling steps, while outlining future pathways to mitigate data scarcity, refine modeling objectives, and expand hybrid modeling applications.