Penalized Cox regression analysis in the high-dimensional and low-sample size settings, with applications to...

Area of Science:

• Bioinformatics
• Genomics
• Biostatistics

Background:

• Microarray technology links gene expression to clinical outcomes, particularly in cancer classification.
• Limited research exists on correlating gene expression with censored survival data (e.g., overall survival, relapse time).
• There is a need for predictive models with high accuracy and parsimony for survival data.

Purpose of the Study:

• To develop a method for selecting genes relevant to patient survival using microarray data.
• To build accurate and parsimonious predictive models for patient survival.
• To address the challenges of high-dimensional and low-sample size data in survival analysis.

Main Methods:

• Utilized L(1) penalized estimation for the Cox model.
• Employed the least-angle regression (LARS) method to efficiently solve computational difficulties.
• Developed the LARS-Cox procedure for gene selection and predictive modeling.

Main Results:

• The LARS-Cox procedure successfully identified genes related to cancer survival.
• Simulation studies and real-world data application demonstrated the method's efficacy.
• LARS-Cox regression outperformed L(2) penalized regression and other dimension-reduction methods in predictive performance.

Conclusions:

• The LARS-Cox procedure is valuable for identifying genes associated with survival phenotypes.
• It enables the creation of parsimonious predictive models for patient survival.
• The method facilitates the classification of future patients into high- and low-risk groups based on gene expression profiles.