Cox-MK: a model-X knockoff framework for genome-wide survival association analysis

In genome-wide survival association studies, time-to-event (TTE) phenotypes are often underutilized due to the challenges of large-scale multiple testing under linkage disequilibrium and heavy censoring. We propose Cox-MK, a novel genome-wide survival analysis framework that integrates knockoff statistics with the saddlepoint approximation (SPA), enabling SNP-level false discovery rate (FDR) control in biobank-scale studies. Simulation studies and real-data applications of UK Biobank data for both common and rare variants demonstrate that the proposed method achieves higher statistical power and well-calibrated FDR control compared with existing approaches. Compared with SPACox, Cox-MK identifies additional candidate genes for TTE traits. Specifically, it detects 47 additional SNPs mapped to 28 genes for asthma and 16 additional SNPs mapped to 12 genes for ischemic heart disease, including CD247 for asthma and EDNRA for ischemic heart disease. Overall, Cox-MK provides an effective tool for prioritizing putative causal variants underlying complex survival phenotypes.