CSMR

CSMR performs component-wise sparse penalized mixture regression and supervised clustering to identify explanatory subspaces linking high-dimensional genetic data and clinical phenotypes.

Key Features:

• Supervised clustering algorithm: Employs a supervised clustering approach that leverages penalized mixture regression models to handle high-dimensional datasets where features (e.g., genetic markers) can exceed sample size.
• Classification Expectation Maximization adaptation: Builds upon the classification expectation maximization framework to enable supervised clustering with improved computational efficiency and biological interpretability.
• Penalized mixture regression model: Incorporates regularization techniques within mixture regression to select relevant feature subspaces that explain component-specific response variables.
• Biological interpretability and computational efficiency: Produces component-specific sparse regression coefficients to facilitate biological interpretation while optimizing computational performance.

Scientific Applications:

• Genetic Studies: Identifies relationships between genetic variations and clinical presentations to dissect molecular bases of disease and support personalized medicine.
• Drug Sensitivity Analysis: Discerns subgroups of cell lines with distinct responses in drug sensitivity datasets to inform pharmacogenomic analyses.

Methodology:

Penalized mixture regression using regularization combined with supervised clustering implemented via a classification expectation maximization framework.