A Method for Heterogeneity Analysis of Complex Diseases Based on Clustering Algorithm

Abstract

There are lots of methods designed for epistasis analysis, but some of them neglect the heterogeneity phenomena of complex diseases. In some cases, the results of association studies may be hard to be interpreted. In this study, we propose a three-step method for heterogeneity analysis. (1) A feature selection step is applied for recognizing multiple combinations of epistatic SNPs which may contribute to different subtypes of complex diseases. (2) A filter step based on Bonferroni-corrected significance level is used to remove those false positive epistatic SNPs combinations. (3) Several clustering algorithms are designed to illustrate and visualize the potential clusters, which are helpful for recognizing the different subtypes of complex diseases. The experimental results show that our method has practical meanings in heterogeneity analysis.