survClust

survClust performs outcome-weighted supervised clustering to integrate somatic mutations, DNA copy number, DNA methylation, mRNA, miRNA, and protein data with survival endpoints (Overall Survival, Progression-Free Survival) to identify prognostic molecular subtypes across TCGA pan-cancer datasets.

Key Features:

• Outcome-Weighted Integration: Leverages survival data (Overall Survival, Progression-Free Survival) to weight molecular features and guide clustering.
• Supervised Clustering: Implements an outcome-weighted supervised clustering algorithm drawing inspiration from supervised text classification techniques.
• Multi-Omics Data Utilization: Processes somatic mutations, DNA copy number variations, DNA methylation, and expression profiles at the mRNA, miRNA, and protein levels.
• Pan-Cancer Application: Applied across 18 cancer types using data from The Cancer Genome Atlas (TCGA).
• Novel Prognostic Subtype Identification: Identifies prognostic subtypes, including groups characterized by high tumor mutation burden with elevated CD8 T cell immune marker expression and aggressive subtypes associated with CDKN2A deletion.
• Visualization Techniques: Incorporates visualization such as circomap to display somatic alterations at genome-wide and individual gene levels.

Scientific Applications:

• Patient Stratification: Identifies survival-associated molecular subtypes for stratifying cancer patients.
• Translational Research: Links molecular profiles to clinical outcomes to support translational research efforts.
• Precision Oncology: Provides prognostic subtype information that can inform personalized treatment strategies and clinical decision-making.

Methodology:

survClust implements an outcome-weighted supervised clustering algorithm that leverages survival endpoints (Overall Survival, Progression-Free Survival), draws inspiration from supervised text classification techniques, and integrates somatic mutations, DNA copy number, DNA methylation, mRNA, miRNA, and protein expression data.