HPV-Adjusted Feature Screening With FDR Control in Head and Neck Cancer

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Human papillomavirus (HPV) significantly impacts head and neck (HN) cancer prognosis, with HPV-positive cases showing better survival.
• Current gene screening methods often ignore HPV status, potentially missing crucial prognostic markers and limiting clinical utility.
• Oversight of HPV status in genomic analyses can confound gene effects, hindering biological interpretation of survival markers.

Purpose of the Study:

• To develop a novel conditional screening method for ultrahigh-dimensional survival data that adjusts for HPV status in HN cancer.
• To identify prognostic genes with independent survival associations while capturing HPV-specific interactions and synergistic effects.
• To enhance the biological interpretability and clinical utility of molecular markers for HN cancer survival.

Main Methods:

• A two-stage, model-free framework combining nonparametric statistics for initial screening.
• A unified false discovery rate (FDR) control procedure for refined feature selection.
• Application to The Cancer Genome Atlas (TCGA) HN cancer data.

Main Results:

• The proposed conditional screening method demonstrates advantages over existing alternatives in simulation studies.
• Analysis of TCGA HN cancer data identified a set of robust prognostic genes.
• New insights into molecular pathways influencing survival across different HPV subgroups were uncovered.

Conclusions:

• The conditional screening framework effectively identifies prognostic genes independent of HPV status.
• The method captures crucial HPV-specific interactions, improving survival prediction in HN cancer.
• This approach offers enhanced biological insights and potential clinical utility for precision oncology.