BJASS

BJASS performs sure joint screening for right-censored, ultra-high-dimensional time-to-event data to identify and retain relevant covariates for sparsity-restricted semiparametric accelerated failure time modeling.

Key Features:

• Sure Joint Screening Procedure: A two-step sparsity-restricted least-squares screening that uses an initial step based on a synthetic time variable and a refinement step using Buckley-James imputed event times, with the refinement step optionally iterated.
• Sparsity-Restricted Semiparametric Accelerated Failure Time Model: Employs a semiparametric accelerated failure time framework with sparsity restriction to accommodate right-censored, ultra-high-dimensional covariates.
• High Retention Probability: For any fixed number of refinement steps, the method retains all significant variables with probability tending to 1.
• Performance Validation: Validated by simulation studies comparing BJASS to marginal screening methods, demonstrating improved identification of relevant covariates.
• Real-World Dataset Applications: Applied to datasets including diffuse large-B-cell lymphoma (DLBCL) and breast cancer data.
• Implementation: Implemented in MATLAB.

Scientific Applications:

• High-dimensional survival analysis: Screening and variable selection for right-censored time-to-event studies with ultra-high-dimensional covariates in biomedical research.
• Disease-specific studies: Application to clinical genomics datasets such as diffuse large-B-cell lymphoma (DLBCL) and breast cancer to identify prognostic covariates.

Methodology:

Initial sparsity-restricted least-squares screening using a synthetic time variable; refinement via sparsity-restricted least-squares on Buckley-James imputed event times with optional multiple iterations; analysis framed within a sparsity-restricted semiparametric accelerated failure time model; implemented in MATLAB.