Genome-wide association studies (GWAS) are able to identify the role of individual SNPs in influencing a phenotype. Nevertheless, such analysis is unable to explain the biological complexity of several diseases. We elaborated an algorithm that starting from genes in molecular pathways implicated in a phenotype is able to identify SNP-SNP interaction's role in association with the phenotype. The algorithm is based on three steps. Firstly, it identifies the biological pathways (gene ontology) in which the genes under analysis play a role (GeneMANIA). Secondly, it identifies the group of SNPs that best fits the phenotype (and covariates) under analysis, not considering individual SNP regression coefficients but fitting the regression for the group itself. Finally, it operates an analysis of SNP interactions for each possible couple of SNPs within the group. The sensitivity and specificity of our algorithm was validated in simulated datasets (HapGen and Simulate Phenotypes programs). The impact on efficiency deriving from changes in the number of SNPs/patients under analysis, linkage disequilibrium and minor allele frequency thresholds was analyzed. Our algorithm showed a strong stability throughout all analysis operated, resulting in an overall sensitivity of 81.67 % and a specificity of 98.35 %. We elaborated a stable algorithm that may detect SNPs interactions, especially those effects that pass undetected in classical GWAS. This method may contribute to face the two relevant limitations of GWAS: lack of biological informative power and amount of time needed for the analysis.
ASJC Scopus subject areas