From Genomes to Ecological Predictions: Using Metabolic Modelling for Uncovering Penicillium expansum Suppression by Native Apple Fruit Epiphytic Bacteria

Background The native carposphere microbiome holds the potential for the biological control of postharvest pathogens. However, traditional biocontrol research relies on empirical, culture-dependent screening, leaving the specific mechanistic interactions within the microbial community unresolved. In this study, a predictive framework integrating metagenomics and genome-scale metabolic modeling was established to translate static genomic data into dynamic, mechanistic predictions of pathogen suppression. The framework was applied to predict the suppression of Penicillium expansum , the causal agent of apple blue mold, by the native epiphytic bacterial community of Malus domestica 'Golden Delicious', followed by co-culture and fruit assay validation. Results 24 high-quality metagenome-assembled genomes (MAGs) were recovered directly from the apple fruit surface. Initial simulations utilizing automated draft metabolic models failed to grow on biologically relevant defined media. Following targeted manual network curation, dynamic simulations were utilized to establish a predictive gradient of pathogen inhibition based on theoretical metabolic resource overlap (MRO). Crucially, a significant positive correlation was observed between these computationally predicted overlaps and the physical restriction of pathogen growth in co-culture assays. For metabolically aggressive strains such as C. terrigena , the models predicted the rapid consumption of shared sugars, robustly validating nutrient sequestration as a primary mode of antagonism. Conversely, the isolate B. nasdae was predicted to be a weak nutrient competitor, yet it exhibited superior biocontrol efficacy in actual fruit wound bioassays. To resolve this discrepancy, computational secretion profiles were analyzed, uniquely delineating B. nasdae as a potent secretor of volatile organic compounds (VOCs). This theoretical alternative mechanism was subsequently validated experimentally, confirming VOC-mediated toxicity as a highly effective, non-metabolic mode of pathogen suppression. Conclusions This study demonstrates that highly curated genome-scale metabolic models can accurately predict pathogen inhibition driven by competitive resource overlaps, while simultaneously providing critical insights into alternative, non-metabolic suppression mechanisms such as volatile toxicity. Furthermore, the failure of automated drafts establishes targeted manual network curation as a prerequisite for utilizing these models to generate accurate ecological predictions. Ultimately, a robust computational baseline for evaluating the specific biocontrol potential of native epiphytes is established.