Detecting disease-related SNP loci based on GSP

Abstract

A large number of studies have shown that susceptibility to some diseases may be related to some SNP (Single Nucleotide Polymorphism) loci. Accurate location of disease-related SNP loci can help people understand the pathogenesis of diseases and prevent them from happening. Based on Graph Signal Processing (GSP) theory, this paper proposes a novel method called GSP to detect disease-related SNP loci. GSP method is divided into five steps. The first step is to establish a graph signal model of all SNP loci. The second step is to extract the high-frequency components of graph signal with a high-pass filter. The third step is to transform the filtered graph signal with graph Fourier transformation. The fourth step is to extract the graph signal corresponding to high frequency with inverse graph Fourier transformation. The fifth step is to transform the extracted signal with Gaussian distribution and use Pauta criterion to screen out disease-related SNP loci. The experimental results show that GSP method can detect all disease-related SNP loci of simulation data and Genetic Disease A (GDA) and three of the four disease-related SNP loci of Age-related Macular Degeneration (AMD) with excellent performance. GSP method provides a new idea for the identification of pathogenicity loci.