HDMAC

HDMAC performs high-dimensional regression analysis to identify molecular alterations associated with survival and binary clinical outcomes in cancer, supporting discovery of prognostic biomarkers and druggable targets.

Key Features:

• Penalized Regression Models: Implements Ridge, Lasso, and Adaptive Lasso penalized regressions applied to Cox proportional hazards (Cox PH) regression for survival outcomes and to logistic regression for binary outcomes.
• First-Step Screening: Incorporates a first-step screening process to reduce the multiple-comparison burden in large genomic datasets and narrow candidate genes for further analysis.
• Model Validation and Prediction Estimation: Reports hazard ratios or estimated coefficients for selected genes, supports construction of multivariate regression models, and uses cross-validation to estimate model predictive power.
• Data Access Integration: Provides R code to download complete sets of molecular variables from The Cancer Genome Atlas (TCGA).

Scientific Applications:

• Ovarian and Bladder Cancer Analysis: Applied to gene mutation and mRNA expression datasets from ovarian and bladder cancer patients to identify candidate genes associated with genetic mutations or abnormal gene expression.

Methodology:

Uses a first-step screening followed by penalized regression (Ridge, Lasso, Adaptive Lasso) applied to Cox PH for survival and logistic regression for binary outcomes; selection yields hazard ratios or estimated coefficients for constructing multivariate models and cross-validation is used to estimate predictive power; includes R code for downloading TCGA molecular variables.