Bayesian Statistics Improves Biological Interpretability of Metabolomics Data from Human Cohorts

Area of Science:

• Metabolomics
• Bayesian Statistics
• Systems Biology

Background:

• Univariate metabolomics analyses typically employ frequentist statistics (p-values) to test hypotheses.
• Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) research often involves complex metabolomic data with potential for subtle differences.
• Frequentist approaches may lack the power to detect significant metabolic alterations in complex diseases.

Purpose of the Study:

• To propose and evaluate the use of Bayesian statistics for analyzing metabolomics data in ME/CFS.
• To demonstrate how Bayesian methods can incorporate prior information from previous studies to improve statistical power.
• To compare Bayesian and frequentist approaches in identifying plasma metabolite differences in ME/CFS patients.

Main Methods:

• Applied Bayesian statistical methods to metabolomics data from three independent human cohorts of ME/CFS patients.
• Utilized results from Study 1 as prior information for analyzing Study 2 data.
• Compared findings with traditional frequentist analyses using Benjamini-Hochberg FDR correction.

Main Results:

• Bayesian analysis identified 97 altered compounds in Study 2 (vs. 0 with frequentist methods) by incorporating Study 1 priors.
• Key findings included altered ether-lipid and triacylglyceride levels, and exposome compounds linked to diet/medication.
• Prostaglandin F2alpha was consistently reduced in ME/CFS patients across all three studies.

Conclusions:

• Bayesian statistics significantly enhances the detection of metabolic alterations in ME/CFS compared to frequentist methods.
• This approach provides superior biological insights and can identify relevant biomarkers.
• Bayesian statistics is recommended for similar metabolomics studies with comparable designs and assays.