α $$ \alpha $$ -KIDS: A Novel Feature Evaluation in the Ultrahigh-Dimensional Right-Censored Setting, With
Atika Farzana Urmi1, Chenlu Ke2, Dipankar Bandyopadhyay1
1Department of Biostatistics, Virginia Commonwealth University, Richmond, Virginia, USA.
We developed alpha-KIDS, a novel method for identifying key genetic markers in head and neck cancer survival data. This approach helps predict patient outcomes and understand cancer progression using genome-wide information.
Area of Science:
- Genomics
- Bioinformatics
- Cancer Research
Background:
- Genome-wide data from sequencing technologies aids head and neck cancer (HNC) diagnosis and prognosis.
- Identifying predictive markers for survival is essential for prognostic systems and understanding cancer drivers.
Purpose of the Study:
- Introduce alpha-KIDS, a model-free feature screening procedure for ultrahigh-dimensional, right-censored survival data.
- Ensure robustness against unknown censoring mechanisms and control the false discovery rate (FDR).
Main Methods:
- A two-stage procedure involving dual screening with nonparametric reproducing-kernel-based ANOVA statistics.
- A unified knockoff procedure for refining feature sets under FDR control.
Main Results:
- Simulation studies demonstrate the finite sample properties and novelty of the alpha-KIDS method.
- Application to The Cancer Genome Atlas (TCGA) head and neck cancer survival data.
Conclusions:
- The alpha-KIDS methodology effectively identifies predictive markers in complex survival data.
- The R package aKIDS is available for implementing this robust feature screening procedure.
More Related Videos
07:13Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
07:15Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
