Key points are not available for this paper at this time.
ABSTRACT Human papillomavirus (HPV) is a well‐established prognostic factor in head and neck (HN) cancer, with HPV‐positive patients exhibiting markedly better survival outcomes compared to their HPV‐negative counterparts. While advances in (cancer) genomics have been pivotal to precision medicine, existing gene screening methods for identifying molecular markers to predict survival often fail to account for HPV status. This oversight can result in missing important genes, whose effects are confounded or overshadowed by HPV, thereby limiting the biological interpretability and clinical utility of identified markers. To address these limitations, we propose a novel conditional screening method for ultrahigh‐dimensional right‐censored survival data that adjusts for HPV status. This approach identifies prognostic genes with independent associations with survival while also capturing HPV‐specific interactions and synergistic effects. The proposed method employs a two‐stage, model‐free framework that combines nonparametric statistics for initial screening with a unified false discovery rate (FDR) control procedure to refine feature selection. Simulation studies demonstrate its advantages over existing alternatives. Application of the conditional screening framework to HN cancer data from The Cancer Genome Atlas revealed a set of robust prognostic genes, uncovering new insights into the molecular pathways driving survival outcomes across HPV subgroups.
Urmi et al. (Sat,) studied this question.