Disease progression based feature screening for ultrahigh-dimensional survival-associated biomarkers

Area of Science:

• Genomics
• Biostatistics
• Cancer Research

Background:

• Ultrahigh-dimensional biomarker data is increasingly available for cancer research.
• Identifying biomarkers related to survival outcomes is crucial for cancer patient management.
• Existing feature screening methods often analyze only one survival endpoint, potentially biasing results.

Purpose of the Study:

• To develop a novel feature screening method for identifying biomarkers related to overall survival (OS) by incorporating time-to-progression (TTP) data.
• To improve the accuracy of biomarker selection for OS by accounting for disease progression.
• To provide a flexible and computationally efficient screening approach.

Main Methods:

• The proposed method ranks features based on their correlation with the conditional distribution of OS given TTP observations.
• It does not require marginal model assumptions for individual endpoints.
• The method is designed for minimal computational cost.

Main Results:

• Theoretical analysis confirms the method's ranking consistency, sure screening, and false discovery rate control.
• Simulations show improved feature selection accuracy compared to methods ignoring TTP.
• The approach demonstrated practical utility in breast cancer genome data analysis for disease classification.

Conclusions:

• Incorporating time-to-progression (TTP) information into feature screening enhances the accuracy of identifying biomarkers for overall survival (OS).
• The proposed method offers a flexible, computationally efficient, and robust approach for biomarker discovery in cancer genomics.
• This technique facilitates more accurate disease classification and improved understanding of survival outcomes in cancer patients.