Synthetic associations are unlikely to account for many common disease genome-wide association signals

Abstract

Synthetic associations have been posited as a possible explanation for missing heritability in complex disease. We show several lines of evidence which suggest that, while possible, these synthetic associations are not common.

Figure 1. SNPs in the NOD2 region as an example of synthetic association.

Left panel represents ten control chromosomes sampled randomly; right panel represents ten random Crohn's disease patient chromosomes. Colored circles represent variant alleles at SNPs genotyped on a GWAS chip, colored explosions represent the three known causal variants in the gene. While none of the GWAS SNPs are strongly correlated with any of the individual causal alleles (the tag SNP theory which underlies the GWAS design), the collective effect of the three causal SNPs is to distort the frequencies of the GWAS SNPs in cases and controls. This collective effect of several low frequency SNPs different distances from a common SNP has been termed a “synthetic association.”

Figure 2. Evidence of association between Crohn's disease and the NOD2 region.

Grey points: results from a logistic regression test of association. Black points: results from a logistic regression test of association after conditioning on compound carrier status for three rare NOD2 mutations (highlighted by red triangles). The NOD2 gene region is denoted by the orange track. The complete eradication of the signal at the common SNPs after conditioning on the rare SNPs demonstrates that the GWAS signal is a synthetic association driven by these rare SNPs.

Figure 3. Comparative power of linkage and GWAS to detect synthetic association.

Solid lines indicate power for the T1DGC linkage analysis (2,658 affected sibpairs), assuming a risk allele frequency of 0.01, and α = 1×10⁻⁴, for N = 3 (black), 5 (red), or 9 (blue) independent risk alleles. Dashed lines indicate power to detect these synthetic associations in a GWAS of 3,000 cases and 3,000 controls (data taken from Dickson et al. Figure 2). Linkage is more powerful than GWAS except for situations with few causal alleles of relatively modest effect.