CBNplot

Workflows

Standard Workflow

Infer and plot a Bayesian network of genes within a specific enriched pathway using expression data.

library(CBNplot)
library(bnlearn)
library(org.Hs.eg.db)

# 1. Load User Inputs & Prepare Data
data(gaussian.test)
kegg <- org.Hs.egPATH2EG
mapped <- mappedkeys(kegg)
genes <- as.list(kegg[mapped])[["00532"]]
counts <- head(gaussian.test, length(genes))
row.names(counts) <- genes

pway <- clusterProfiler::enrichKEGG(gene = genes)
pway <- clusterProfiler::setReadable(pway, org.Hs.eg.db, keyType="ENTREZID")

# 2. Visualize the gene-level Bayesian network for the first pathway
bngeneplot(results = pway, exp = counts, pathNum = 1, expRow="ENTREZID")

# 3. Extract the underlying network structure, strength, and direction
ret <- bngeneplot(results = pway, exp = counts, pathNum = 1, returnNet=TRUE, expRow="ENTREZID")
head(ret$str)

# 4. Convert the averaged network to an igraph object and perform centrality analysis
g <- bnlearn::as.igraph(ret$av)
igraph::evcent(g)$vector

# 5. Customize the gene network visualization
bngeneplotCustom(results = pway, exp = counts, expRow="ENTREZID", pathNum=1, fontFamily="sans", glowEdgeNum=NULL, hub=1)

Inputs/Outputs: Takes an enrichment analysis result object and a normalized expression data frame as inputs, and outputs Bayesian network plots and network structure data frames.

When to Use

• To infer and visualize Bayesian networks of genes within a specific enriched pathway using expression data via bngeneplot.
• To visualize relationships between multiple pathways as a Bayesian network using bnpathplot.
• To customize network visualizations with glow edges or hub highlighting using bngeneplotCustom or bnpathplotCustom.
• To extract network strength and direction or convert the inferred network to an igraph object using bnlearn::as.igraph for centrality analysis.

When NOT to Use

• For general non-pathway-based gene regulatory network inference without prior enrichment results (use packages like minet or GENIE3 instead).
• For static pathway diagrams without expression-based network inference (use pathview instead).

Data Requirements

• Enrichment analysis results (e.g., from clusterProfiler::enrichKEGG or ReactomePA).
• A normalized expression matrix or data frame (e.g., counts) where row names match the gene identifier type (e.g., ENTREZID) specified in the enrichment results.

Key Parameters

• results: Enrichment analysis result object (e.g., from clusterProfiler::enrichKEGG).
• exp: Normalized expression data frame or matrix.
• pathNum (1): The pathway number in the enrichment results to plot.
• expRow ("ENTREZID"): The column/row identifier type for expression data.
• returnNet (FALSE): Logical indicating whether to return the network structure, strength, and direction.
• nCategory (5): Number of pathways to include in bnpathplot.
• glowEdgeNum (NULL): Parameter in custom plots to highlight edges of interest.
• hub (1): Parameter in custom plots to highlight hub nodes.

Best Practices

• Ensure gene identifiers in the expression matrix match the keyType of the enrichment results (e.g., convert using clusterProfiler::setReadable).
• Use mappedkeys to map pathway identifiers to genes when preparing test or custom pathway gene lists.
• Set returnNet = TRUE in bngeneplot to inspect the underlying network strength and direction before relying solely on the visual plot.

Common Pitfalls

• Mismatched gene identifier types between expression data and enrichment results. Fix: Use clusterProfiler::setReadable to align key types (e.g., to "ENTREZID").
• Providing raw, unnormalized count data to network inference. Fix: Supply normalized expression values (e.g., log-transformed or normalized counts).

Alternatives

• pathview for overlaying expression data onto static KEGG pathway diagrams.
• enrichplot for standard enrichment visualizations (e.g., cnetplot, emapplot) without Bayesian network inference.
• bnlearn for general Bayesian network inference without direct integration with enrichment results.

Citations

• Scutari M (2010). Learning Bayesian Networks with the bnlearn R Package. Journal of Statistical Software.
• Yu G, Wang LG, Han Y, He QY (2012). clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters. OMICS.

References

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