compound.Cox

compound.Cox implements univariate feature selection and multigene predictor construction in an R package to relate gene expression data to patient survival outcomes.

Key Features:

• Univariate Feature Selection: Performs univariate significance tests (Wald tests or score tests) to identify genes associated with survival and selects an optimal subset for model development.
• Cross-Validation Algorithm: Applies cross-validation to assess the predictive capability and robustness of gene-based predictors.
• Permutation Algorithm: Implements permutation testing to evaluate false discovery rate (FDR) and assess the significance of gene selection.
• Multigene Predictor Construction: Constructs multigene predictors using compound covariate, compound shrinkage, and copula-based algorithms that integrate genes identified by univariate selection.

Scientific Applications:

• Survival Analysis: Predicts patient survival from gene expression profiles in oncology studies, including lung cancer.
• Gene Selection Optimization: Determines optimal significance levels for univariate tests to refine selection of predictive genes.
• False Discovery Rate Computation: Computes and evaluates FDR to reduce spurious associations in gene selection.

Methodology:

Performs univariate significance tests to select genes, applies cross-validation and permutation testing to refine models and assess statistical validity, and constructs multigene predictors using compound covariate, compound shrinkage, or copula-based algorithms.