False Discovery Rate Control in Cancer Biomarker Selection Using Knockoffs

Area of Science:

• Genomics
• Biostatistics
• Oncology

Background:

• Accurate cancer risk assessment, diagnosis, prognosis, and treatment prediction rely on identifying informative biomarkers.
• High-throughput genomics enables biomarker selection from numerous genes, but controlling false discoveries is difficult due to large gene numbers and small patient cohorts.
• Existing methods struggle with the complexity of gene-disease associations and diverse patient data.

Purpose of the Study:

• To develop and validate a robust method for unbiased biomarker discovery in cancer genomics.
• To control the rate of false discoveries in high-dimensional genomic studies.
• To identify genes associated with clinical outcomes in skin cutaneous melanoma.

Main Methods:

• Utilized a knockoff procedure for controlling false discoveries in genomic data.
• Developed a flexible method accommodating various covariate distributions, linear/nonlinear associations, and survival models.
• Validated the method through simulations and application to real clinical data.

Main Results:

• The proposed knockoff procedure demonstrated superior performance compared to alternative methods in simulations.
• The method successfully identified seven significant genes associated with Breslow thickness in skin cutaneous melanoma.
• The findings highlight the utility of the knockoff procedure for reliable biomarker discovery in oncology.

Conclusions:

• The knockoff procedure offers a reliable approach to control false discoveries in cancer biomarker identification.
• This method enhances the accuracy of genomic biomarker selection for clinical applications.
• The identified genes provide potential targets for understanding and managing skin cutaneous melanoma.