GMHCC

GMHCC performs graph-based multiple hierarchical consensus clustering to partition high-throughput biomolecular data, including cancer gene expression and single-cell RNA-seq, into robust subpopulations for downstream analyses such as differential gene, gene ontology, and KEGG pathway analysis.

Key Features:

• Graph-based Unsupervised Feature Ranking (FR): Constructs a graph over pairwise features and assigns ranks to select relevant biomolecular data attributes.
• Diverse Basic Partitions (BPs) and Hierarchical Structures: Generates multiple diverse feature subsets to produce several basic partitions and refines them by exploring hierarchical structures.
• Global Consensus Function: Integrates basic partitions into a consensus by leveraging hierarchical relationships among partitions.
• Graph-based Linking Method: Explicitly models inter-cluster relationships via a graph-based linking approach to produce a final partition that preserves cluster integrity and connectivity.

Scientific Applications:

• Validation on benchmark datasets: Evaluated on 35 cancer gene expression datasets and eight single-cell RNA-seq datasets, showing superior performance compared to several state-of-the-art consensus clustering methods.
• Differential Gene Analysis: Facilitates identification of differentially expressed genes across inferred clusters.
• Gene Ontology and KEGG Pathway Analysis: Supports downstream gene ontology enrichment and KEGG pathway analysis to interpret biological processes and pathways associated with clusters.
• Subtype discovery and lineage characterization: Enables discovery of molecular subtypes and exploration of cellular developmental lineages and characterization mechanisms.

Methodology:

Constructs a graph over pairwise features for unsupervised feature ranking; generates multiple diverse feature subsets to create basic partitions; refines partitions via a global consensus function that explores hierarchical structures; applies a graph-based linking method to consider inter-cluster relationships and produce the final partition.