Information fusion of CNVs and SNPs on gene-gene interactions for molecular subtypes of lymphoma

Although genome-wide association studies report many disease-associated loci involved in pathogenesis, current identified variants only explain a little part of the heritability underlying complex diseases. To explore the other missing part of heritability, data-mining methods are proposed and developed to detect disease-associated interactions between variants. Recently, some studies have revealed the linkage disequilibrium between chromosome structure variations and disease-associated loci. We are motivated to employ a fusion approach that incorporates the information of copy number variations (CNVs) for identifying interactions between single nucleotide polymorphisms (SNPs). The CNV profiles are first used for clustering analysis of disease subtypes, and then the SNP-SNP interactions are examined by the multifactor dimensionality reduction (MDR) method. We applied the fusion approach in analyzing 214 lymphoma cases. The results showed that the interactions identified by the fusion approach were more significantly associated with lymphoma than those identified only by MDR without incorporating CNV information. Therefore, we conclude that information fusion of CNVs and SNPs provides a proper strategy for detecting gene-gene interactions in disease association studies.