Variable selection with Group LASSO approach: Application to Cox regression with frailty model

Area of Science:

• Biostatistics
• Genomics
• Epidemiology

Background:

• Traditional Cox proportional hazards models struggle with high-dimensional gene expression data and large covariate numbers relative to sample size.
• Predicting survival outcomes solely on individual gene expression is insufficient due to complex biological pathway regulation.
• The Cox model's homogeneity assumption is often violated in real-world populations with unmeasured risk factors.

Purpose of the Study:

• To propose a novel statistical method, group LASSO with gamma-distributed frailty, for variable selection in Cox regression.
• To address limitations of traditional models in analyzing survival outcomes with high-dimensional clinical and gene expression data.
• To account for heterogeneity among population groups based on exposure and susceptibility.

Main Methods:

• Developed group LASSO with gamma-distributed frailty for variable selection in Cox regression.
• Extended existing methodologies to incorporate group structures and account for population heterogeneity.
• Established the consistency property of the proposed statistical method.

Main Results:

• The proposed group LASSO method demonstrated promising performance in simulated and real-world data analyses.
• Outperformed other methods such as group SCAD and group MCP in variable selection accuracy.
• The method is suitable for diverse research areas, including genetics and environmental health (e.g., air pollution).

Conclusions:

• Group LASSO with gamma-distributed frailty offers a robust approach for survival outcome analysis with complex, high-dimensional data.
• The method effectively handles population heterogeneity and the interplay of multiple biological factors.
• Future research may explore extensions using adaptive group LASSO and sparse group LASSO for further advancements.