The contribution of rare variation to prostate cancer heritability

Although genome-wide association studies (GWAS) have found more than a hundred common susceptibility alleles for prostate cancer, the GWAS reported variants jointly explain only 33% of risk to siblings, leaving the majority of the familial risk unexplained. We use targeted sequencing of 63 known GWAS risk regions in 9,237 men from four ancestries (African, Latino, Japanese, and European) to explore the role of low-frequency variation in risk for prostate cancer. We find that the sequenced variants explain significantly more of the variance in the trait than the known GWAS variants, thus showing that part of the missing familial risk lies in poorly tagged causal variants at known risk regions. We report evidence for genetic heterogeneity in SNP effect sizes across different ancestries. We also partition heritability by minor allele frequency (MAF) spectrum using variance components methods, and find that a large fraction of heritability (0.12, s.e. 0.05; 95% CI [0.03, 0.21]) is explained by rare variants (MAF<0.01) in men of African ancestry. We use the heritability attributable to rare variants to estimate the coupling between selection and allelic effects at 0.48 (95% CI of [0.19, 0.78]) under the Eyre-Walker model. These results imply that natural selection has driven down the frequency of many prostate cancer risk alleles over evolutionary history. Overall our results show that a substantial fraction of the risk for prostate cancer in men of African ancestry lies in rare variants at known risk loci and suggests that rare variants make a significant contribution to heritability of common traits.