BioNAR

Workflows

Standard Workflow

Build, annotate, and analyze a protein-protein interaction (PPI) network to identify central proteins, functional communities, bridging proteins, and disease/annotation overlaps.

library(BioNAR)

# 1. Build the network from a data frame
file <- system.file("extdata", "PPI_Presynaptic.csv", package = "BioNAR")
tbl <- read.csv(file, sep="\t")
gg <- buildNetwork(tbl)

# 2. Annotate vertices with gene names and diseases
gg <- annotateGeneNames(gg)
afile <- system.file("extdata", "flatfile_human_gene2HDO.csv", package = "BioNAR")
dis <- read.table(afile, sep="\t", skip=1, header=FALSE, strip.white=TRUE, quote="")
gg <- annotateTopOntoOVG(gg, dis)

# 3. Estimate vertex centrality measures
gg <- calcCentrality(gg)
mc <- getCentralityMatrix(gg)

# 4. Fit degree distribution to power law and estimate entropy rate
pFit <- fitDegree(as.vector(igraph::degree(graph=gg)), threads=1, Nsim=5, plot=FALSE)
ent <- getEntropyRate(gg)
SRprime <- getEntropy(gg, maxSr = NULL)

# 5. Perform community detection (clustering) and evaluate modularity
nm <- normModularity(gg, alg = 'louvain')
mem <- calcMembership(gg, alg = 'louvain')
gg <- calcClustering(gg, alg = 'louvain')

# 6. Recluster dense communities
remem <- calcReclusterMatrix(gg, mem, alg = 'louvain', 10)

# 7. Build consensus matrix and evaluate cluster robustness
conmat <- makeConsensusMatrix(gg, N = 5, alg = 'louvain', type = 2, mask = 10, reclust = FALSE)
clrob <- getRobustness(gg, alg = 'louvain', conmat)

# 8. Calculate and plot bridgeness
br <- getBridgeness(gg, alg = 'louvain', conmat)
gg <- calcBridgeness(gg, alg = 'louvain', conmat)
g <- plotBridgeness(gg, alg = 'louvain', VIPs = c('8495', '22999'), Xatt = 'SL')

Input: A data frame of protein-protein interactions. Output: An annotated igraph object with calculated centralities, community memberships, and bridgeness metrics.

When to Use

• Analyzing protein-protein interaction (PPI) networks to identify key topological features and central proteins using calcCentrality().
• Detecting functional communities or complexes using multiple clustering algorithms (e.g., Louvain, Walktrap) via calcAllClustering().
• Identifying bridging proteins that connect different functional communities using getBridgeness().
• Evaluating network scale-free properties and entropy rates using fitDegree() and getEntropyRate().

When NOT to Use

• For reconstructing networks directly from raw gene expression data (e.g., co-expression network construction); use WGCNA instead.
• For basic, non-biological general graph theory operations where standard igraph or tidygraph suffices without biological annotations.

Data Requirements

• Input data frame with interaction pairs (columns representing interacting nodes, e.g., Entrez IDs).
• Annotation files mapping Entrez IDs to gene names, diseases (HDO), or GO terms.

Key Parameters

• alg: Clustering algorithm to use (e.g., "louvain", "wt", "fc", "infomap", "lec").
• N: Number of randomization rounds for consensus matrix generation (typically 500).
• type: Sampling scheme for consensus matrix (1 for sampling edges, 2 for sampling vertices).
• mask: Percentage of edges or vertices to mask during perturbation.
• Xatt: Centrality measure to plot against bridgeness (e.g., "SL" for semilocal centrality).

Best Practices

• Ensure all nodes have non-empty GeneName attributes before running downstream annotation and analysis functions.
• Use calcAllClustering() to compare multiple community detection algorithms and evaluate their modularity using clusteringSummary().
• Run at least 500 randomization rounds (N = 500) in makeConsensusMatrix() for real-world robustness analysis.

Common Pitfalls

• Using too few randomization rounds for consensus matrix: Leads to inaccurate bridgeness and robustness estimates. Fix: Increase N to at least 500 in makeConsensusMatrix().
• Missing gene name annotations: Causes downstream functions that rely on gene names to fail. Fix: Run annotateGeneNames() and verify with any(is.na(V(gg)$GeneName)).

Alternatives

• igraph: For general network analysis and basic clustering without specialized biological workflows.
• pRoloc: For spatial proteomics and organelle assignment (not network-topology based).

Citations

• Menche et al. 2015, Science (for disease-disease overlap).
• Nepusz et al. 2008 (for bridgeness).

References

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