Zenith

Workflows

Standard Workflow

Perform gene set enrichment analysis on differential expression results from a linear mixed model fit with dream and visualize the results.

library(zenith)
library(edgeR)
library(variancePartition)
library(tweeDEseqCountData)
library(kableExtra)

# 1. Filter low-expressed genes
data(pickrell)
geneCounts = exprs(pickrell.eset)
df_metadata = pData(pickrell.eset)
dsgn = model.matrix(~ gender, df_metadata)
keep = filterByExpr(geneCounts, dsgn, min.count=5)

# 2. Compute library size normalization
dge = DGEList(counts = geneCounts[keep,])
dge = calcNormFactors(dge)

# 3. Estimate precision weights
vobj = voomWithDreamWeights(dge, ~ gender, df_metadata)

# 4. Fit linear (mixed) model
fit = dream(vobj, ~ gender, df_metadata)
fit = eBayes(fit)

# 5. Load gene sets and run zenith
msdb.gs = get_MSigDB("H", to="ENSEMBL")
res.gsa = zenith_gsa(fit, msdb.gs, 'gendermale', progressbar=FALSE)
plotZenithResults(res.gsa)

Input: Raw RNA-seq counts and sample metadata. Output: A data frame of gene set enrichment statistics and a heatmap visualization of the top results.

When to Use

• To perform gene set enrichment analysis on differential expression results generated from linear (mixed) models using variancePartition::dream().
• To account for correlation between gene expression traits by extending the camera method from limma to linear mixed models.
• To easily fetch and cache gene sets from MSigDB (using get_MSigDB) or Gene Ontology (using get_GeneOntology).

When NOT to Use

• For simple fixed-effect designs without random effects or sample correlation; standard limma::camera is sufficient.
• If you are not using variancePartition::dream for differential expression; zenith_gsa is specifically designed to take the output of dream().

Data Requirements

• A fitted model object from variancePartition::dream() that has been processed with limma::eBayes().
• A GeneSetCollection object containing gene sets (e.g., loaded via get_MSigDB or get_GeneOntology).
• Gene identifiers in the gene sets must match the rownames of the fitted model (e.g., both using ENSEMBL or SYMBOL).

Key Parameters

• to ("ENSEMBL"): Target gene identifier type (e.g., "ENSEMBL", "SYMBOL", "ENTREZ") in get_MSigDB or get_GeneOntology.
• progressbar (FALSE): Logical indicating whether to show a progress bar during zenith_gsa execution.
• coef ('gendermale'): The coefficient/contrast in the fitted model to evaluate for gene set enrichment.

Best Practices

• Filter out low-expressed genes using edgeR::filterByExpr to improve statistical power and model stability.
• Estimate precision weights using variancePartition::voomWithDreamWeights to account for the mean-variance relationship in RNA-seq count data.
• Use plotZenithResults to visualize the top gene set enrichment results across coefficients or cell types.

Common Pitfalls

• Gene identifier mismatch: Ensure the gene ID type specified in get_MSigDB(..., to=...) matches the ID type used as rownames in the expression matrix and the dream fit.
• Running on raw counts: Always normalize counts using calcNormFactors and apply voomWithDreamWeights before fitting models with dream.

Alternatives

• limma::camera for fixed-effect linear models.
• EnrichmentBrowser::getGenesets for loading alternative gene set databases like KEGG or Enrichr.

Citations

• Hoffman G (2026). Zenith: Gene set analysis for linear mixed models. R package.
• Wu D, Smyth GK (2012). Camera: a competitive gene set test accounting for inter-gene correlation. Nucleic Acids Research.

References

• Homepage: bioconductor.org/packages/zenith
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/zenith/inst/doc/zenith.html