Gene‐Informed Modeling of Denitrification Process in Groundwater Through Dynamic Flux Balance Analysis and Deep Learning

Abstract We introduce two gene‐informed approaches that, for the first time, explicitly incorporate microbial genomic information into groundwater denitrification simulations. First, we couple the dynamic flux balance analysis (DFBA, which resolves genome‐resolved metabolic networks) with the PFLOTRAN reactive transport model (RTM). The ensuing DFBA‐reactive transport model (RTM) framework provides a mechanistic, pathway‐explicit linkage between gene expression and solute transformation dynamics. Second, we develop a gene‐informed deep‐learning model that leverages environmental covariates and functional gene abundances to emulate biogeochemical reaction kinetics at orders‐of‐magnitude reduced computational cost. Using controlled batch and column experiments including alternating surface‐water/groundwater flow regimes, both models reproduce observed NO 3 − depletion and NO 2 − transients and outperform the conventional RTM constrained only by geochemistry. Our DFBA‐RTM elucidates pathway‐level controls on denitrification, whereas the gene‐informed deep‐learning model offers fast and accurate forecasts suitable for multi‐scenario analyses. Together, these results show that incorporating microbial genomic information substantially improves hydro‐biogeochemical modeling results and that mechanistic and data‐driven strategies are complementary. While the former supports process attribution, the latter yields efficient and scalable surrogates. The framework is readily extensible to other contaminant and redox systems, advancing hydro‐biogeochemical modeling toward genetically informed (high‐resolution) representations of subsurface environments.