ConsensusClusterPlus

Dependencies & Environment

Package-intrinsic requirements from the Bioconductor landing page — reproduce in any R environment.

• Version: 1.76.0 · Bioconductor: 3.23 · R: ≥ 4.6
• Depends: ALL, cluster
• Imports: Biobase, ALL, cluster
• System requirements: URL
• Install: BiocManager::install("ConsensusClusterPlus")

When to Use

• Unsupervised class discovery from gene expression data (e.g., microarray data).
• Determining optimal cluster count and membership using consensus clustering resampling.
• Generating consensus matrices, CDF plots, and tracking plots to assess cluster stability.
• Calculating cluster-consensus and item-consensus metrics using the calcICL function.

When NOT to Use

• For extremely large datasets without pre-computing distance matrices (use a pre-computed dist object instead to save time).
• If you need non-resampling based clustering (use standard hclust or kmeans instead).

Data Requirements

• A matrix of numerical values where columns are samples (items) and rows are features (e.g., genes).
• Alternatively, a pre-computed distance matrix (dist object) can be provided.
• Data is typically filtered for highly variable features (e.g., top 5000 genes by mad) and median-centered using sweep and median.

Key Parameters

• d: The input data matrix or pre-computed distance matrix.
• maxK: The maximum evaluated cluster count (e.g., 6).
• reps: The number of resamplings to perform (e.g., 50 for testing, 1000 in practice).
• pItem: The proportion of items to resample in each iteration (e.g., 0.8).
• pFeature: The proportion of features to resample in each iteration (e.g., 1).
• clusterAlg: The clustering algorithm to use (e.g., "hc", "pam", "km", or a custom function).
• distance: The distance metric to use (e.g., "pearson", "spearman", "euclidean", "manhattan").
• plot: The output format for graphical results (e.g., "png", "pdf", "pngBMP").

Best Practices

• Filter the dataset to the most informative/variable genes using mad before clustering.
• Median-center the data using sweep and median if using Pearson correlation distance.
• Use a high number of reps (e.g., 1,000) and a higher maxK (e.g., 20) in real-world practice.
• Run calcICL after ConsensusClusterPlus to generate and plot cluster-consensus and item-consensus metrics.

Common Pitfalls

• Computation time is too long for large datasets with feature resampling. Fix: Provide a pre-computed distance matrix to d or use plot="pngBMP" to avoid plotting massive dendrograms.
• Results change between runs making analysis irreproducible. Fix: Always specify a random seed parameter in the ConsensusClusterPlus call.
• Consensus matrix dendrogram fails to plot for large datasets. Fix: Use plot="pngBMP" which uses the bitmap function rather than png.

Alternatives

• cluster: For standard clustering algorithms like diana or pam without consensus resampling.
• NMF: For non-negative matrix factorization based clustering.
• mclust: For model-based clustering.

Citations

• Monti, S., Tamayo, P., Mesirov, J., Golub, T. (2003) Consensus Clustering: A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data. Machine Learning, 52, 91-118.
• Wilkerson, M.D., Hayes, D.N. (2010). ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics, 26(12):1572-3.

References

• Homepage: https://bioconductor.org/packages/ConsensusClusterPlus
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/ConsensusClusterPlus/inst/doc/ConsensusClusterPlus.pdf

<!-- biomate-cta -->

---

Run this on BioMate

This skill is the knowledge layer — when, why, and how to use consensusclusterplus. To run this analysis on your own data with managed compute, automated QC, and reproducible outputs, use BioMate — free to start.

▶ Open consensusclusterplus on BioMate →