HD-GMCM

HD-GMCM models high-dimensional biological data using Gaussian mixture copulas to decouple marginal distributions from inter-variable dependencies for improved clustering and patient subtyping.

Key Features:

• Decoupling Marginals and Dependencies: Uses copulas to separate marginal distribution modeling from dependency structure modeling, enabling flexible representation of complex dependencies.
• High-Dimensional Scalability: Overcomes parameter-inference challenges of traditional copula-based methods to scale to high-dimensional genomic and clinical datasets.
• Robustness to Non-Gaussian Data: Handles non-Gaussian marginal distributions and increases resilience to outliers via Gaussian mixture copulas.
• Improved Clustering Performance: Captures latent structures in high-dimensional gene-expression and clinical datasets to yield superior model-based clustering results.
• Interpretable Subtype Characterization: Explicit modeling of dependencies permits interpretation of clusters and characterization of patient subtypes.

Scientific Applications:

• Precision Medicine: Facilitates accurate patient subtyping to inform tailored therapeutic strategies.
• Oncology — Lung cancer (TCGA): Applied to lung cancer data from The Cancer Genome Atlas (TCGA) as a case study demonstrating practical utility in oncology research.
• Clinical Subtype Discovery: Supports identification of clinically relevant subtypes for personalized treatment planning and survival analysis.

Methodology:

HD-GMCM employs Gaussian mixture copulas to model marginal distributions and dependency structures independently for model-based clustering in high-dimensional datasets.