Enrichment visualization

Version Compatibility

Reference examples tested with: ggplot2 3.5+

Before using code patterns, verify installed versions match. If versions differ:

• R: packageVersion('<pkg>') then ?function_name to verify parameters

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Enrichment Visualization

"Create publication-quality plots from my enrichment analysis" → Generate dotplots, gene-concept networks, enrichment maps, GSEA running score plots, and ridgeplots from clusterProfiler results.

• R: dotplot(), cnetplot(), emapplot(), gseaplot2() (enrichplot)

Scope

This skill covers enrichplot package functions designed for clusterProfiler results:

• dotplot(), barplot() - Summary views
• cnetplot(), emapplot(), treeplot() - Network/hierarchical views
• gseaplot2(), ridgeplot() - GSEA-specific
• goplot(), heatplot(), upsetplot() - Specialized views

For custom ggplot2 dotplots and statistical annotation, see data-visualization/distribution-plots and data-visualization/ggplot2-fundamentals.

Setup

Goal: Load required packages for visualizing enrichment analysis results.

Approach: Import clusterProfiler, enrichplot, and ggplot2 which provide the plotting functions for enrichment objects.

library(clusterProfiler)
library(enrichplot)
library(ggplot2)

# Assume ego (enrichGO result), kk (enrichKEGG result), or gse (GSEA result) exists

Dot Plot

Goal: Summarize enrichment results showing gene ratio, count, and significance in a single figure.

Approach: Use enrichplot dotplot which maps gene ratio to x-axis, term to y-axis, dot size to count, and color to p-value.

Most common visualization - shows gene ratio, count, and significance.

dotplot(ego, showCategory = 20)

# Customize
dotplot(ego, showCategory = 15, font.size = 10, title = 'GO Enrichment') +
    scale_color_gradient(low = 'red', high = 'blue')

# Save
pdf('go_dotplot.pdf', width = 10, height = 8)
dotplot(ego, showCategory = 20)
dev.off()

Bar Plot

Shows enrichment count or gene ratio.

barplot(ego, showCategory = 20)

# Customize
barplot(ego, showCategory = 15, x = 'GeneRatio', color = 'p.adjust')

Gene-Concept Network (cnetplot)

Goal: Visualize which genes contribute to multiple enriched terms, revealing shared biology.

Approach: Build a bipartite network connecting enriched terms to their member genes, optionally colored by fold change.

Shows relationships between genes and enriched terms.

# Basic cnetplot
cnetplot(ego)

# With fold change colors
cnetplot(ego, foldChange = gene_list)

# Circular layout
cnetplot(ego, circular = TRUE, colorEdge = TRUE)

# Customize node size
cnetplot(ego, node_label = 'gene', cex_label_gene = 0.8)

Enrichment Map (emapplot)

Goal: Identify clusters of related enriched terms by visualizing shared gene overlap.

Approach: Compute pairwise term similarity, then plot as a network where edges connect terms sharing genes.

Shows term-term relationships based on shared genes.

# Requires pairwise_termsim first
ego_pt <- pairwise_termsim(ego)
emapplot(ego_pt)

# Customize
emapplot(ego_pt, showCategory = 30, cex_label_category = 0.6)

# Cluster by similarity
emapplot(ego_pt, group_category = TRUE, group_legend = TRUE)

pairwise_termsim() Method Selection

# Default: Jaccard Coefficient (works with any gene set type)
ego_pt <- pairwise_termsim(ego)

# For GO terms: Wang semantic similarity (more biologically meaningful)
ego_pt <- pairwise_termsim(ego, method = 'Wang', semData = godata('org.Hs.eg.db', ont = 'BP'))

Method	Type	When to Use
JC (Jaccard)	Gene overlap	Default; works with KEGG, Reactome, any gene set
Wang	Graph-based	Best for GO; captures biological relationships independent of annotation version
Resnik/Lin/Jiang	IC-based	GO only; depends on annotation corpus (results change between database releases)

Tree Plot

Hierarchical clustering of enriched terms.

ego_pt <- pairwise_termsim(ego)
treeplot(ego_pt)

# Show more categories
treeplot(ego_pt, showCategory = 30)

Upset Plot

Show overlapping genes between terms.

upsetplot(ego)

# Limit to specific number of terms
upsetplot(ego, n = 10)

GSEA-Specific Plots

Running Score Plot (gseaplot2)

# Single gene set
gseaplot2(gse, geneSetID = 1, title = gse$Description[1])

# Multiple gene sets
gseaplot2(gse, geneSetID = 1:3)

# With subplots
gseaplot2(gse, geneSetID = 1, subplots = 1:3)

# By term ID
gseaplot2(gse, geneSetID = 'GO:0006955')

Ridge Plot

Distribution of fold changes in gene sets.

ridgeplot(gse)

# Top n gene sets
ridgeplot(gse, showCategory = 15)

# Order by NES
ridgeplot(gse, showCategory = 20) + theme(axis.text.y = element_text(size = 8))

Reading ridge plots:

• Shifted right (positive values): Gene set enriched among upregulated genes
• Shifted left (negative values): Gene set enriched among downregulated genes
• Bimodal distribution: Pathway contains both strongly up- and down-regulated genes; may indicate heterogeneous pathway with opposing components
• Narrow peak: Enrichment driven by a small cluster of similarly ranked genes
• Broad distribution: Many genes with varied rankings (more diffuse, less concentrated signal)

GO-Specific Plot (goplot)

DAG structure of GO terms.

# Only for GO enrichment results
goplot(ego)

# Specific ontology
goplot(ego_bp)  # where ego_bp is enrichGO with ont='BP'

Heatplot

Gene-concept heatmap.

heatplot(ego, foldChange = gene_list)

# Customize
heatplot(ego, showCategory = 15, foldChange = gene_list)

Compare Multiple Analyses

Goal: Visualize enrichment results side by side across multiple gene lists or conditions.

Approach: Use dotplot on compareCluster output, optionally faceting by cluster.

# Compare clusters (from compareCluster)
dotplot(ck, showCategory = 10)

# Facet by cluster
dotplot(ck) + facet_grid(~Cluster)

Customize ggplot2 Elements

Goal: Fine-tune enrichment plots with custom titles, themes, colors, and text sizes.

Approach: Chain ggplot2 modifiers onto enrichplot output since all functions return ggplot2 objects.

All enrichplot functions return ggplot2 objects.

p <- dotplot(ego, showCategory = 20)

# Add title
p + ggtitle('GO Biological Process Enrichment')

# Change theme
p + theme_minimal()

# Adjust text
p + theme(axis.text.y = element_text(size = 10))

# Change colors
p + scale_color_viridis_c()

Save Plots

Goal: Export enrichment plots as publication-quality PDF or PNG files.

Approach: Use base R pdf/png device functions or ggplot2 ggsave to write plots to files.

# PDF (vector, publication quality)
pdf('enrichment_plots.pdf', width = 10, height = 8)
dotplot(ego, showCategory = 20)
dev.off()

# PNG (raster)
png('dotplot.png', width = 800, height = 600, res = 100)
dotplot(ego, showCategory = 20)
dev.off()

# Using ggsave
p <- dotplot(ego)
ggsave('dotplot.pdf', p, width = 10, height = 8)

Visualization Summary

Function	Best For	Input Type
dotplot	Overview of enrichment	ORA, GSEA
barplot	Simple counts/ratios	ORA
cnetplot	Gene-term relationships	ORA
emapplot	Term clustering	ORA
treeplot	Hierarchical grouping	ORA
upsetplot	Term overlap	ORA
gseaplot2	Running enrichment score	GSEA
ridgeplot	Fold change distribution	GSEA
goplot	GO DAG structure	GO only
heatplot	Gene-concept matrix	ORA

Choosing the Right Visualization

Goal	Plot	Key Tip
First overview of top enriched terms	dotplot	Best starting point; shows 3 dimensions (ratio, count, p-value)
Which genes drive multiple enriched terms	cnetplot	Limit to 5-10 terms; use circular = TRUE for crowded networks
Identify functional modules among terms	emapplot	Run pairwise_termsim() first; if everything connects to everything, results are redundant
GSEA: detailed single-pathway view	gseaplot2	Check where genes cluster in the ranked list
GSEA: overview of all enriched sets	ridgeplot	Read direction (left/right shift) and shape (narrow vs broad)
Compare enrichment across conditions	dotplot on compareCluster	Use facet_grid(~Cluster) for side-by-side panels

Common Visualization Mistakes

• Too many terms: plots with > 30 terms are unreadable. Use showCategory = 15-20.
• Not simplifying GO first: showing 15 redundant GO terms (cell cycle, cell cycle process, mitotic cell cycle...) wastes space and misleads. Run simplify() before plotting.
• Missing gene set size: always show both the overlap count and the total pathway size. A 3/5 overlap (60%) is very different from 30/500 (6%).
• Bar plots for GSEA: bar plots show count or enrichment. For GSEA, use NES on the x-axis, not p-value. Use dotplot or ridgeplot instead.
• Skipping pairwise_termsim(): emapplot and treeplot will fail or produce meaningless results without it.

Related Skills

• go-enrichment - Generate GO enrichment results
• kegg-pathways - Generate KEGG enrichment results
• gsea - Generate GSEA results