AI-Assisted Spatial Metabolic Engineering in Plants: Integrating Flux Design, Spatial Omics, and Synthetic Biology

Background: Plant synthetic biology reprograms metabolic networks for the sustainable production of high-value compounds. Recent computational advances incorporate machine learning to accelerate the design-build-test-learn (DBTL) cycle, enabling more predictable and scalable engineering in photoautotrophic chassis. However, the translation of AI-generated designs into stable plant phenotypes remains constrained by incomplete plant-specific training datasets, tissue heterogeneity, and limited in vivo validation. Scope: This review examines the convergence of machine learning methods with plant metabolic engineering across four spatial engineering levels: subcellular compartmentalization, cell/tissue/organ-specific control, developmental or inducible regulation, and genome-level organization. Spatial omics is considered a cross-cutting validation layer, and the evidence supporting each technology is classified as plant-demonstrated, non-plant proof-of-concept, or prospective. Conclusions: Integrating predictive machine learning with spatial engineering offers promising strategies to design complex biosynthetic pathways. Hybrid approaches, combining constraint-based metabolic models with generative algorithms, reduce trial-and-error in crop engineering. Future plant synthetic biology is likely to rely increasingly on automated and data-rich workflows to support more predictable plant bioproduction.