Employing weighted biological network structure for finding disease genetic markers in SNP association studies

SNP association study has been widely performed to find disease-related genetic markers usually by investigating the difference of SNP genotype frequencies between disease and non-disease samples and evaluating its significance in a statistical sense. However, such approach often incurs the problem of producing tie scores over multiple SNPs, especially when the number of samples is not large enough. In addition, for the finding of genetic markers related to complex diseases, such as cancer, various environmental or functional factors need to be taken into consideration. Thus, to deal with these problems, we examine a new analytical framework to the identification of disease-related genetic markers that can incorporate a variety of a priori known gene functional factors into SNP association study. Specifically, using any biological resources of interest, we build up a gene network structure and gives high ranks to SNPs associated with hub genes (or core genes), which have high connectivity to many other genes, in determining disease markers. For experiments, we constructed network structure with gene ontology annotations and cancer modules, respectively, and identified genetic markers related to prostate cancer and non-Hodgkin lymphoma, respectively. The results demonstrated that the use of network structure constructed with available various biological resources can lead to the better finding of disease-related genetic markers in SNP association studies.