Predicting Strain-Specific Metabolic Capabilities in the Genus Pseudomonas with a Flux-to-AI Approach Reveals Hidden Cell Envelope Properties

Abstract

Bacteria from the Pseudomonas genus are omnipresent in air, soil, and water. They have been widely studied for their broad metabolic versatility and their presence in epidemiological chains, bioproduction, bioremediation, and disease processes. Each year, more genomic sequences are reported in databases and repositories. However, the relationship between genomic variability in Pseudomonas strains and the diversity of their metabolic capabilities across environmentally relevant phenotypes remains unknown. Additionally, predictive tools for analyzing different strains within a systematic framework are limited. Here, we reconstructed genome-scale metabolic models (GEMs) of 44 Pseudomonas strains from various environments and investigated their capabilities to metabolize different carbon sources and metabolic intermediaries. By systematically testing the substrate utilization of the models, we demonstrate how GEM-predicted capabilities can differentiate between strains and that high metabolic versatility is associated with the predicted ability of the strains to remove toxic compounds while maintaining core functionalities. Hundreds of model simulations were used as input to a machine-learning classification schema, resulting in the identification of metabolic capabilities that better differentiate species. Interestingly, transcription and expression models supported these findings by showing that the metabolic capabilities identified by the Flux-to-AI framework are associated with differences in proteome allocation across strains, including pathways involved in amino acid and carbohydrate metabolism, ultimately revealing strain-specific resource allocation strategies linked to cell-envelope-associated functions.