Detecting survival-associated biomarkers from heterogeneous populations

Area of Science:

• Genomics
• Biostatistics
• Cancer Research

Background:

• Genomic biomarkers are crucial for predicting patient survival and guiding cancer treatment.
• High-throughput technologies generate abundant genomic data, but identifying reliable biomarkers across studies is challenging due to cohort heterogeneity.
• Existing variable selection methods are limited to single-study analyses, hindering the discovery of robust biomarkers.

Purpose of the Study:

• To propose a novel method, Cox-TOTEM, for detecting survival-associated biomarkers common across multiple genomic studies.
• To address the challenges of heterogeneity and weak signals in multi-study genomic data analysis.
• To identify reproducible and clinically applicable prognostic factors.

Main Methods:

• Developed a two-stage variable selection method based on the Cox proportional hazards model, named Cox-TOTEM.
• Applied Cox-TOTEM to analyze multiple genomic studies, focusing on detecting shared survival-associated biomarkers.
• Utilized simulations to evaluate the method's performance compared to existing single-study approaches.

Main Results:

• Cox-TOTEM significantly improved the sensitivity of variable selection compared to analyzing studies separately.
• The method demonstrated superior performance, particularly with weak biological signals and heterogeneous study populations.
• Application to TCGA transcriptomic data identified key survival-associated genes in Pan-Gynecologic cancers.

Conclusions:

• Cox-TOTEM offers a robust approach for identifying common survival biomarkers from multiple heterogeneous genomic studies.
• The method enhances the discovery of reproducible biomarkers, facilitating better clinical decision-making in cancer.
• Identified essential survival genes linked to common mechanisms in Pan-Gynecologic cancers, advancing our understanding of disease progression.