Table 3 Overview of the research questions, corresponding methodological approaches, selected models, and performance metrics used in this benchmark
From: A systematic benchmark of integrative strategies for microbiome-metabolome data
Scientific question | Research aim | Model selection | Selected models | Performance metrics | |
|---|---|---|---|---|---|
Is there any relationship between microorganisms and metabolites at a global level? | Global associations | Need to adjust for covariates | YES: Regression-based models | MMiRKAT | Type-I error rate, Power |
NO: Bi-directional models | Mantel test, Procrustes analysis | ||||
Are microbiome and metabolome datasets summarizable through a limited number of components? | Data summarization | Does the directionality matter | YES: Regression-based models | PLS-Regression RDA | % Explained Variance |
NO: Canonical-based models | PLS-Canonical CCA MOFA2 | ||||
Can we identify associations between metabolites and species? | Individual associations | Need to adjust for covariates | YES: Regression-based models | Log-Contrast MiRKAT CLR-LM | Type-I error rate, Power |
NO: Correlation-based models | Pearson Spearman HALLA | ||||
Can we identify core microorganisms and metabolites? | Feature selection | Need to account for the between-within correlation | YES: Univariate models | CODA-LASSO CLR-(M)LASSO | Sparsity sensitivity specificity |
NO: Multivariate-models | sPLS-Regression sPLS-Canonical sCCA | ||||
