Performance of model-based multifactor dimensionality reduction methods for epistasis detection by controlling population structure

Abstract Background In genome-wide association studies the extent and impact of confounding due to population structure have been well recognized. Inadequate handling such is likely lead spurious associations, hampering replication, identification causal variants. Several strategies developed for protecting associations against confounding, most popular one based on Principal Component Analysis. contrast, in gene-gene interaction epistasis are much less investigated understood. particular, role nonlinear genetic substructure detection largely under-investigated, especially outside a regression framework. Methods To identify variants synergy, improve interpretability replicability results, we introduce three model-based multifactor dimensionality reduction approach structured populations, namely MBMDR-PC, MBMDR-PG, MBMDR-GC. Results Simulation results comparing performance various approaches show that presence MBMDR-PC MBMDR-PG consistently better control type I error rate at nominal level than Moreover, our proposed methods correction outperform MDR-SP terms statistical power. Conclusion We demonstrate through extensive simulation effect degrees relatedness propose appropriate remedial measures linear sample similarity.