hopach

hopach constructs hierarchical cluster trees by recursively partitioning datasets, ordering and optionally collapsing clusters using the Mean/Median Split Silhouette (MSS) criterion, and estimating element-to-cluster membership probabilities via non-parametric bootstrap for applications in genomics and molecular biology.

Key Features:

• Recursive partitioning: Constructs hierarchical trees of clusters by recursively partitioning the input dataset.
• Cluster ordering and collapse: Orders clusters at each tree level and can collapse clusters to simplify the hierarchy.
• Mean/Median Split Silhouette (MSS): Uses the MSS criterion to identify the tree level that contains maximally homogeneous clusters.
• Final ordered list: Runs the hierarchical tree down to produce a final ordered list of elements.
• Non-parametric bootstrap: Estimates the probability that each element belongs to each cluster using non-parametric bootstrap resampling.

Scientific Applications:

• Genomics: Segmentation and clustering of high-throughput genomics datasets to identify homogeneous groups of features or samples.
• Molecular biology: Analysis of molecular biology data to organize features or samples into ordered clusters with statistical membership estimates.
• High-throughput data analysis: Interpretation and organization of large-scale expression or other high-throughput assay data into hierarchical, ordered clusters.

Methodology:

Recursive partitioning to build hierarchical trees, ordering and optional collapse of clusters, selection of optimal tree level using the Mean/Median Split Silhouette (MSS) criterion, running the tree down to produce a final ordered list, and non-parametric bootstrap to estimate element-to-cluster membership probabilities.