Predicting metabolic fluxes from omics data via machine learning: Moving from knowledge-driven towards data-driven...

Area of Science:

• Systems biology
• Computational biology
• Metabolic engineering

Background:

• Accurate prediction of microbial phenotypes is crucial for systems biology.
• Genome-scale models (GEMs) with constraint-based methods like FBA predict metabolic fluxes but require extensive prior knowledge.
• Integrating omics data (transcriptomics, proteomics) to enhance phenotype prediction accuracy remains a challenge.

Purpose of the Study:

• To develop and evaluate a novel machine learning (ML) approach for predicting metabolic fluxes under various conditions.
• To compare the performance of omics-based ML models against the established parsimonious FBA (pFBA) method.
• To assess the utility of integrating transcriptomics and/or proteomics data for improved phenotype prediction.

Main Methods:

• Development of supervised machine learning models utilizing transcriptomics and/or proteomics data.
• Application of the ML approach to case studies involving *Escherichia coli*.
• Comparative analysis of ML model predictions against predictions from the pFBA method.

Main Results:

• The proposed omics-based ML approach demonstrated promising results for predicting metabolic fluxes.
• ML models achieved smaller prediction errors for both internal and external metabolic fluxes compared to pFBA.
• The study provides a viable alternative for metabolic flux prediction, especially when integrating omics data.

Conclusions:

• Omics-data-driven machine learning offers a powerful strategy for enhancing the accuracy of metabolic flux predictions in microbial systems.
• This approach addresses limitations of traditional methods by leveraging experimental omics data.
• The developed methodology holds significant potential for advancing systems biology and metabolic engineering applications.