Making a plan for metabolomics analysis

Analytical workflow

Before I jumped into analysis, I surveyed some recent metabolomics papers to see if I could come up with a consensus on analytical method:

Huffmyer et al. (2023):

• PLS-DA in mixOmics package
• WCNA with WGCNA correlated to identify with coral developmental time points
• analyzed for enrichment of compound class and KEGG pathway with MetaboAnalyst web interface

Williams et al. (2021):

• DGCA for correlation of metabolite pairs between ambient and stressed Mcap
• Significant correlations —> used to create co-occurrence networks

Noisette et al. (2021):

• PCA then PLS-DA
• VIP identified from PLS-DA
• ANOVA for differences in metabolites between groups
• functional enrichment and pathway topology analysis (takes into account role of metabolite, position, and direction of interaction in a network)
• MetaboAnalyst used for all analyses

Trigg et al. (2019):

• All analyses in R except for PLS-DA and random forest analyses in MetaboAnalyst
• Remove compounds detected in less than half of crabs prior to downstream analyses
• univariate analyses: two-way ANOVA for each compound (OA + DO), Benjamini-Hochberg FDR correction. sig compounds (model ANOVA < 0.01 and effect < 0.05 w/o FDR correction) used for pathway analyses
• multivariate: normalized by mean-centering and scaling, PLS-DA with default settings
• Metamapp: map metabolite relationships

Looks like a PLS-DA and some sort of MetaboAnalyst enrichment are common place, so I’ll start there! Huffmyer et al. (2023) has the most recent scripts available on Github, so I’ll base a lot of my work off of that.

Going forward

1. Analyze metabolomics data