Abstract

Eukaryotic metabolism is organized into subcellular compartments enclosed by lipid-membranes. Since most metabolites do not move freely across the membranes, the compartmentalization influences metabolic pathway connectivities. Thus, compartmentalizing the pathways also in genome-scale metabolic models (GEMs) is essential for accurate predictions of eukaryotic cells metabolic phenotypes. Compartmentalization has manually been introduced into the model Eukaryote GEMs like for yeast Saccharomyces cerevisiae. Non-model organisms GEMs can be automatically reconstructed from genome data. However, the existing GEM reconstruction methods do not introduce compartmentalization into the models. To that end, we integrated our novel deep learning protein localization prediction and protein functional annotation into a top-down GEM reconstruction for automatically creating species-specific compartmentalized GEMs. We developed also a universal fungal GEM for top-down reconstruction of species-specifically compartmentalized GEMs and reconstructed models for 834 species. The novel protein localization prediction outperformed the state of the art in classifying proteins into multiple compartments and integrating the enzyme localization predictions to functional annotations improved the top-down model reconstruction. Interestingly, the clustering of the fungal GEMs reconstructed with predicted species-specific enzyme localization resembled more closely the phylogenetic relationships of the species than that of the GEMS without the compartmentalization. Compartmentalization of metabolism in fungi differs from other eukaryotes but the enzyme subcellular localizations vary also among fungal species. The Fungal kingdom encompasses species important for human health, environment, and industrial biotechnology. The reconstructed fungal GEM set offers valuable tools for e.g., predicting metabolic phenotypes of e.g., mushrooms or single-cellular fungi, the roles of eukaryotes in microbial communities, designs optimizing eukaryotic hosts for industrial chemical production, and identifying drug targets against pathogenic fungi. Beyond fungi, the compartmentalized GEM reconstruction method allows combining other universal GEMs for developing cell type specific GEMs for higher eukaryotes including plants, insects, and mammals.