Simulated metabolic profiles reveal biases in pathway analysis methods

Area of Science:

• Computational Biology
• Systems Biology
• Metabolomics

Background:

• Pathway analysis (PA) methods, initially for transcriptomics, can introduce biases in metabolomics.
• Exometabolomics data presents unique challenges due to the distance between measured metabolites and internal disruptions.
• Experimental validation of PA method performance is difficult when the true metabolic state is unknown.

Purpose of the Study:

• To demonstrate that PA can lead to non-specific enrichment in metabolomics.
• To highlight the potential for false assumptions about the causes of perturbed metabolic states.

Main Methods:

• Utilized in silico metabolic modeling to create defined disruptions in metabolic networks.
• Employed the SAMBA (constraint-based modeling) approach to simulate metabolic profiles for pathway knockouts.
• Assessed the ability of PA methods to identify the known disrupted pathway from simulated profiles.

Main Results:

• Complete pathway blockages were not always significantly enriched in simulated metabolomics profiles.
• Enrichment failures were observed despite known disruption sites.
• Factors influencing enrichment include the PA method, pathway set definition, and network structure.

Conclusions:

• Certain metabolomics data may not be suitable for standard PA methods.
• This study provides a benchmark for evaluating and improving PA tools.
• Highlights the need for new PA tools tailored for metabolomics data analysis.