Reconstruction of Eukaryotic Compartmentalized Genome-Scale Metabolic Models Using Deep Learning for over 700 Fungi | AIChE

Reconstruction of Eukaryotic Compartmentalized Genome-Scale Metabolic Models Using Deep Learning for over 700 Fungi

Authors 

Castillo, S. - Presenter, VTT Technical Research Centre of Finland
Blomberg, P., VTT Technical Research Centre of Finland
Jouhten, P., EMBL Heidelberg
Eukaryotic metabolism is organized into sub-cellular compartments. The confined compartments allow diversified environments for adjusting enzyme activities and determine metabolic pathway connections. The pathway maps described as genome-scale metabolic models allow simulations of cellular metabolism such as predictions of metabolic capabilities, inter-species interactions, and drug targets. Such models have been manually curated for model organisms and for other species they can be reconstructed from genome data. However, the compartmentalization defining metabolic pathway connections poses a challenge for automatic reconstruction of models for eukaryotic cells. The protein features allowing eukaryotic cells to sort them in their sub-cellular localizations have so far not been used for creating species-specific compartmentalization in eukaryotic models.

We integrated protein localization prediction using a novel deep learning method and functional annotation to top down genome-scale metabolic model reconstruction for creating species-specific compartmentalized genome-scale metabolic models. We further developed an annotated universal fungal metabolic model for top down reconstruction of species-specifically compartmentalized models for 765 fungal species. The universal fungal model provides a platform for community development.

Fungal kingdom encompasses species important for human health, environment, and industrial biotechnology. The reconstructed model set offers valuable hypothesis generation tools for understanding the role of eukaryotic microbes in microbial communities and metabolic capabilities of mushrooms, designing optimized eukaryotic hosts for industrial biotechnology, and identifying drug targets against pathogenic fungi. Beyond fungi, integrating protein localization prediction to model reconstruction allows combining other universal models for reconstructing cell type specific models for other eukaryotes including plants, insects, and mammals.