B1460
Title: Feature evaluation for ultrahigh dimensional survival data with applications to head and neck cancers
Authors: Chenlu Ke - Virginia Commonwealth University (United States) [presenting]
Abstract: Head and neck cancer ranks as the 6th most prevalent cancer worldwide, with an anticipated 1.08 million new cases annually. Advances in sequencing technologies have allowed the collection of massive genome-wide information that substantially enhances the diagnosis and prognosis of head and neck cancer. Identifying predictive markers for survival outcomes is one of the crucial tasks for devising prognostic systems and learning the underlying molecular driver of the cancer course. A novel, model-free feature evaluation procedures are developed for ultrahigh dimensional survival analysis, notable for their robustness against unknown censoring mechanisms and heavy censoring. The efficacy of the proposed method is justified in theory and its advantages are demonstrated over existing alternatives with numerical studies. Applications to head and neck cancer data result in an independent prognostic signature that successfully differentiates low-risk and high-risk patients in the cancer genome atlas cohort and an external validation cohort.
