Consensus Hierarchical Clustering (ConsensusClusterPlus)

Overview

Consensus hierarchical clustering (implemented in the R/Bioconductor package ConsensusClusterPlus) is an unsupervised subgroup discovery method that repeatedly subsamples items and features, applies agglomerative hierarchical clustering to each subsample, and accumulates a consensus matrix measuring the proportion of subsamples in which each pair of items clusters together. The optimal number of clusters k is selected by inspecting the area under the CDF of consensus matrix values. Distance metrics and linkage methods are user-specified; Spearman correlation distance with complete linkage is commonly used for transcriptome data.

Used by

  • Applied to RNA-seq data (Spearman distance, k=3) for 28 metastatic neuroendocrine neoplasms (pog570_bcgsc_2020); defined three transcriptome clusters: Cluster A (small-intestinal NETs/PanNETs with MEN1/DAXX/ATRX alterations), Cluster B (high-grade/MYC-driven mixed primary sites), and Cluster C (PulNETs/MTCs); NECs and NET-G3s did not form a distinct cluster PMID:24326773.
  • Applied in single-cell RNA-seq analysis (Deng M, cited) to classify cholangiocarcinoma into BA-active and BA-inactive metabolic subtypes; BA-active subtype associated with shorter OS and immunotherapy resistance PMID:25608663
  • Consensus hierarchical clustering applied to top-1,500 variant genes in 329 melanoma RNA-seq samples, defining three transcriptomic subclasses (Immune 51%, Keratin 31%, MITF-low 18%) with distinct survival and mutational profiles. PMID:26091043
  • Used to define three ILC transcriptional subtypes (reactive-like, immune-related, proliferative) from 1,277 SAM-differentiating genes (q=0) in n=106 LumA ILC, with survival differences validated in METABRIC PMID:26451490
  • Consensus hierarchical clustering applied to 1,300 tumor-specific CpG probes from merged HM27 + HM450 methylation data to define six pan-glioma methylation subtypes (LGm1–6) in 932 TCGA glioma samples PMID:26824661
  • Used for sub-group discovery across miRNA, mRNA, and methylation data in 40-66 MRT cases; yielded 2 miRNA sub-groups, 2 mRNA sub-groups (recapitulating AT/RT vs RTK distinction), and 2 WGBS methylation sub-groups (correlated with age >1 year at diagnosis). PMID:26977886
  • Consensus hierarchical clustering used for mRNA-based subtype discovery across 173 PCPG tumors; yielded four subtypes (kinase signaling, pseudohypoxia, Wnt-altered, cortical admixture) validated in an independent Burnichon et al. cohort PMID:28162975.
  • Used for NMF consensus clustering of 408 BLCA RNA-seq samples defining five mRNA expression subtypes (luminal-papillary 35%, luminal-infiltrated 19%, luminal 6%, basal-squamous 35%, neuronal 5%) with distinct survival (p=4×10⁻⁴) PMID:28988769

Notes

  • ConsensusClusterPlus requires a pre-selected feature set (e.g., top-variable genes); results can be sensitive to feature selection cutoff.
  • Visualization typically accompanies the consensus matrix with heatmaps and silhouette plots.
  • Related methods include NMF-based consensus clustering and k-means consensus clustering.

Sources

This page was processed by crosslinker on 2026-05-14. - PMID:25608663

This page was processed by crosslinker on 2026-05-14. - PMID:26091043

This page was processed by crosslinker on 2026-05-14. - PMID:26451490

This page was processed by crosslinker on 2026-05-14. - PMID:26824661

This page was processed by entity-page-writer on 2026-05-15. - PMID:26977886

This page was processed by entity-page-writer on 2026-05-15. - PMID:28162975

This page was processed by wiki-cli on 2026-05-14. - PMID:28988769

This page was processed by entity-page-writer on 2026-05-15.