GDHC

GDHC implements hierarchical clustering based on a nonparametric measure of general dependence to detect linear and nonlinear relationships in high-throughput biological data such as gene expression arrays and liquid chromatography-mass spectrometry (LC-MS) datasets for discovery of regulatory patterns.

Key Features:

• General dependence measure: Uses a sensitive nonparametric measure of general dependence between groups of random variables in high-dimensional spaces.
• Linear and nonlinear capture: Captures both linear and nonlinear relationships among features, including genes and metabolites measured on continuous scales.
• High-dimensional robustness: Designed to operate on high-dimensional, noisy datasets containing thousands of features.
• Supported data types: Applicable to gene expression arrays and liquid chromatography-mass spectrometry (LC-MS) datasets.
• Validation: Simulation studies demonstrate improved identification of nonlinear dependencies compared to hierarchical clustering methods based on correlation and mutual information.
• Implementation: Implemented in R.

Scientific Applications:

• Gene expression clustering: Clustering of gene expression arrays to identify groups of co-dependent genes.
• Metabolomics analysis: Clustering of LC-MS metabolomics data to identify dependent metabolite groups.
• Time-series microarrays: Analysis of microarray datasets measuring gene expression across cell-cycle time series.
• Regulatory relationship detection: Detection of nonlinear dependencies among genes or metabolites to infer regulatory patterns.

Methodology:

Performs hierarchical clustering using a sensitive nonparametric general-dependence measure between groups of random variables in high-dimensional spaces, validated by simulation comparisons to correlation- and mutual information-based hierarchical clustering, and implemented in R.