Gene‐Gene Interaction Analysis for the Survival Phenotype Based on the Kaplan‐Meier Median Estimate
In this study, we propose a simple and computationally efficient method based on the multifactor dimensional reduction algorithm to identify gene‐gene interactions associated with the survival phenotype. The proposed method, referred to as KM‐MDR, uses the Kaplan‐Meier median survival time as a classifier. The KM‐MDR method classifies multilocus genotypes into a binary attribute for high‐ or low‐risk groups using median survival time and replaces balanced accuracy with log‐rank test statistics as a score to determine the best model. Through intensive simulation studies, we compared the power of KM‐MDR with that of Surv‐MDR, Cox‐MDR, and AFT‐MDR. It was found that KM‐MDR has a similar power to that of Surv‐MDR, with less computing time, and has comparable power to that of Cox‐MDR and AFT‐MDR, even when there is a covariate effect. Furthermore, we apply KM‐MDR to a real dataset of ovarian cancer patients from The Cancer Genome Atlas (TCGA).
