MiCARA

MiCARA (Microbiome Confounder-Adjusted Robust Association Analysis) provides an end-to-end framework for quality control, metadata imputation, differential abundance testing, cross-domain residual correlation analysis, and interactive Sankey network visualization.

Recommended Workflow

MiCARA follows a streamlined 5-step pipeline:

1. validate_inputs() — check data formatting, reconstructs raw counts from relative abundances/proportions if needed and filters out high-missingness covariates (>30%).
2. impute_metadata() (Optional) — Imputes remaining missing metadata values (<30%). > When to skip: Skip if your metadata is 100% complete or if your study protocol requires Complete Case Analysis (CCA).
3. run_diff_abundance() — Performs ANCOM-BC2 differential abundance on taxa and pathways.
4. compute_residualCLR_correlations() — Infers residual cross-domain correlations while adjusting for disease covariates.
5. compute_global_fdr() (Optional) global multiple-testing correction

By default, MiCARA identifies significant taxon–pathway associations using disease-specific Benjamini–Hochberg correction in compute_residualCLR_correlations().

For analyses involving multiple disease groups, users may optionally apply an additional global Benjamini–Hochberg correction across all tested taxon–pathway pairs using compute_global_fdr()

This adds globally adjusted q_global values to the $all_links tables and stores the globally significant associations in $global_links, while leaving the original $links results unchanged.

Global correction is optional and is not required for the standard MiCARA workflow.

6. plot_interaction_sankey() — Renders interactive Sankey network diagrams for significant interactions.

Installation

You can install MiCARA directly from GitHub using pak:

# Install pak if not already available
if (!requireNamespace("pak", quietly = TRUE)) {
    install.packages("pak")
}

# Install MiCARA
pak::pak("BushnaqSafa/MiCARA")

Quickstart Example

library(MiCARA)

# 1. Validate & filter inputs
obj <- validate_inputs(
    taxa_mat = taxa_mat,
    pathway_mat = pathway_mat,
    metadata = metadata,
    disease_col = "disease"
)

# 2. Impute missing metadata (optional)
obj <- impute_metadata(obj)

# 3. Perform differential abundance on taxa and pathways
diff_res <- run_diff_abundance(obj, feature_type = c("taxa", "pathways"))

# 4. Infer layer-2 residual cross-domain correlations
interactions <- compute_residualCLR_correlations(obj, diffab_obj = diff_res)

# 5. Optional global multiple-testing correction
interactions_global <- compute_global_fdr(interactions)

# 6. Visualize interactions in a Sankey network
plot_interaction_sankey(interactions, disease = "UC", direction = "both")

# 7. To plot globally significant associations:

plot_interaction_sankey(
    interactions_global,
    disease = "IBD",
    significance = "global"
)