Evaluation of metabolite-microbe correlation detection methods.

Area of Science:

• Microbiome research
• Metabolomics
• Bioinformatics

Background:

• Microbiome data analysis requires specialized correlation methods due to compositional structures and sequencing variations.
• Discovering biological associations between microbes and host metabolites is a growing area of interest in omics studies.

Purpose of the Study:

• To comprehensively evaluate six different correlation methods for detecting microbe-metabolite associations.
• To identify the most effective statistical methods for cross-omics correlation analysis.

Main Methods:

• Evaluated Pearson correlation, Spearman correlation, Sparse Correlations for Compositional data (SparCC), Correlation inference for Compositional data through Lasso (CCLasso), Mutual Information Coefficient (MIC), and Cosine similarity.
• Utilized three simulated and two real-world datasets (public and in-house) to assess method performance.

Main Results:

• Each method demonstrated unique strengths and weaknesses across different scenarios.
• Spearman correlation and MIC exhibited superior overall performance in detecting microbe-metabolite correlations.
• Performance was assessed based on specificity, sensitivity, similarity, accuracy, and stability under varying sparsity levels.

Conclusions:

• Spearman correlation and MIC are recommended for robust microbe-metabolite association analysis.
• The study provides strategic guidance for selecting appropriate correlation methods in multi-omics research.
• Effective correlation detection is crucial for advancing our understanding of host-microbiome interactions.