Robust Model Selection and Estimation for Censored Survival Data with High Dimensional Genomic Covariates

Area of Science:

• Genomics
• Biostatistics
• Survival Analysis

Background:

• Analyzing genomic data for survival outcomes faces challenges: censored data, high dimensionality, and non-normal distributions.
• Existing methods may struggle with these complex data characteristics.

Purpose of the Study:

• To propose a robust method for detecting significant genes related to survival outcomes.
• To simultaneously address censored outcomes, high-dimensional genomic data, and data non-normality.

Main Methods:

• Utilized an Accelerated Failure Time (AFT) model incorporating a general loss function.
• Employed model regularization, shrinkage techniques, and parameter tuning for robust estimation.
• Developed an Expectation-Maximization algorithm for efficient implementation.

Main Results:

• Simulation studies showed the proposed method outperforms existing approaches, especially with heavy-tailed errors and correlated covariates.
• The method provides robust estimation for identifying genes associated with survival.

Conclusions:

• The novel AFT-based method effectively handles complex challenges in genomic survival data analysis.
• The approach offers a powerful tool for identifying significant genes and improving survival outcome predictions.