GraphGDel: Constructing and Learning Graph Representations of Genome-Scale Metabolic Models for Growth-Coupled Gene...

Area of Science:

• Metabolic Engineering
• Computational Biology
• Systems Biology

Background:

• Genome-scale metabolic models are crucial for optimizing cellular functions.
• Current gene deletion strategies often overlook the inherent network structure of these models.
• Developing effective computational approaches for predicting gene deletion strategies remains a challenge.

Purpose of the Study:

• To develop a novel computational framework for predicting gene deletion strategies in metabolic models.
• To leverage graph representations and deep learning for improved prediction accuracy.
• To enable simultaneous cell growth and target metabolite production.

Main Methods:

• A systematic pipeline was created to construct graph representations from constraint-based metabolic models.
• A deep learning framework was developed integrating graph and sequence data.
• The framework was evaluated on three distinct metabolic models.

Main Results:

• The proposed graph-based deep learning approach significantly outperformed existing baselines.
• Accuracy improvements ranged from 5.10% to 16.26% compared to various established methods.
• The method demonstrated consistent performance across different metabolic models.

Conclusions:

• Graph representations combined with deep learning offer a powerful approach for predicting gene deletion strategies.
• This work advances the field of metabolic engineering by providing a more accurate computational tool.
• The developed framework has the potential to optimize bioproduction processes.