polyClustR

polyClustR reconciles clustering solutions from multiple algorithms to identify robust molecular subtype communities in high-dimensional biological datasets, supporting molecular stratification in cancer research.

Key Features:

• Reconciliation of clustering solutions: Integrates results from diverse clustering algorithms into subtype "communities" using statistical approaches such as the hypergeometric test and measures of relative sample proportions.
• Systematic consensus clustering: Applies multiple consensus clustering algorithms and systematically reconciles their outputs to enhance identification of consistent subtypes.
• Validation across datasets: Demonstrated on a breast cancer dataset and applied to uveal melanoma datasets, identifying subtype communities associated with metastasis-free survival differences and known chromosomal aberrations.

Scientific Applications:

• Cancer molecular subtyping: Defines molecular subtypes for patient stratification and treatment optimization in oncology.
• Prognostic association discovery: Identifies subtype communities associated with metastasis-free survival and chromosomal aberrations.
• Cross-study subtype harmonization: Reconciles disparate clustering results across datasets to enable robust comparisons of molecular subtypes.

Methodology:

Integration of clustering results from multiple clustering and consensus clustering algorithms, with statistical reconciliation using hypergeometric tests and measures of relative sample proportions to define subtype "communities".