A Novel Approach to Detect Cumulative Genetic Effects and Genetic Interactions in Crohn’s Disease

Abstract

Genome-wide association studies have identified at least 71 Crohn's disease (CD) genetic risk loci, but the role of gene-gene interactions is unclear. The value of genetic variants in clinical practice is not defined because of limited explained heritability. We examined model predictability of combining the 71 CD risk alleles and genetic interactions in an ongoing inflammatory bowel disease genome-wide association study. The Wellcome Trust Case Control Consortium inflammatory bowel disease genome-wide association study was used as a replicate cohort. We used logic regression, an adaptive regression methodology, to search for high-order binary predictors (e.g., single-nucleotide polymorphism [SNP] interactions). The combined 71 CD SNPs had good CD risk predictability (area under the curve of 0.75 and 0.73 in the 2 cohorts). Higher cumulative allele score predicted higher CD risk, but a relatively small difference in cumulative allele scores was observed between CD and controls (49 versus 47, P < 0.001). Through LR, we identified high-order genetic interactions and significantly improved the model predictability (area under the curve, from 0.75 to 0.77, P < 0.0001). A genetic interaction model, including NOD2, ATG16L1, IL10/IL19, C13orf31, and chr21q loci, was discovered and successfully replicated in the independent Wellcome Trust Case Control Consortium cohort. The explained heritability of the 71 CD SNPs alone was 24% and increased to 27% after adding the genetic interactions. A novel approach allowed the identification and replication of genetic interactions among NOD2, ATG16L1, IL10/IL19, C13orf31, and chr21q loci. CD risk can be predicted by a model of 71 CD loci and improved by adding genetic interactions.