Background Genome-wide association studies have been successful in identifying association between several common variants and coronary artery disease (CAD). However, collectively these variants explain only a small proportion of CAD heritability. It is becoming increasingly clear that the remainder of the “missing CAD heritability” could be explained by low frequency/rare alleles. Because of the small number of observations for any given rare allele, the power to detect its association with a phenotype is a major limiting factor in genetic analysis. In this study we have undertaken a novel statistical approach that combines information from all low frequency (MAF<5%) SNPs at one locus in gene-centric analysis of CAD. We hypothesised that patients with CAD will show over-representation of rare alleles compared to controls.

Methods To examine associations between rare alleles and CAD, we have used data from 2119 CAD cases and 2440 healthy controls recruited to the Welcome Trust Case-Control Consortium (WTCCC) Study. DNA from each subject was genotyped for approximately 45 000 SNPs in more than 2000 genes/loci using 50K IBC array (version 1). Association analysis was based on the CCRaVAT (Case-Control Rare Variant Analysis Tool) algorithm that maximises statistical power by combining all rare alleles within defined regions into a single “super locus”. Differences in the proportion of cases and controls carrying rare “super loci” were tested by Pearson's or Fisher's exact test. Empirical p values were generated by permuting case-control status a predefined number of times and repeating the analysis for each replicate.

Results 5 candidate regions (MMP23B, VEGFA, DVL1, RIPK1, LPAL2) showed an over-accumulation of rare alleles in patients with CAD when compared to controls (FDR<50%). The number of analysed rare alleles at each of these loci ranged from 4 to 42. The most significant over-representation of rare variants were identified at MMP23B (matrix metallopeptidase 23B gene; p=1.3/10⁴), a gene previously unsuspected to play a major role in CAD and VEGFA (vascular endothelial growth factor A; p=2.6×10⁻⁴). Only one of the identified genes (LPAL2; p=1.7×10⁻³) lies within the locus that was previously shown to harbour rare variants associated with susceptibility to CAD.

Conclusions Rare alleles are associated with predisposition to CAD and this gene-centric analysis combining information from low-frequency variants of the same locus has a potential to uncover, at least a proportion of, the “missing heritability” of CAD.