Genome-wide association studies (GWAS) have revealed that the genetic contribution to certain complex diseases is well-described by Fisher’s infinitesimal model in which a vast number of polymorphisms each confer a small effect. Under Fisher’s model, variants have additive effects both across loci and within loci. However, the latter assumption is at odds with the common observation of dominant or recessive rare alleles responsible for monogenic disorders. Here, we searched for evidence of non-additive (dominant or recessive) effects for GWAS variants known to confer susceptibility to the highly heritable quantitative trait, refractive error. Of 146 GWAS variants examined in a discovery sample of 228,423 individuals whose refractive error phenotype was inferred from their age-of-onset of spectacle wear, only 8 had even nominal evidence (p < 0.05) of non-additive effects. In a replication sample of 73,577 individuals who underwent direct assessment of refractive error, 1 of these 8 variants had robust independent evidence of non-additive effects (rs7829127 within ZMAT4, p = 4.76E−05) while a further 2 had suggestive evidence (rs35337422 in RD3L, p = 7.21E−03 and rs12193446 in LAMA2, p = 2.57E−02). Accounting for non-additive effects had minimal impact on the accuracy of a polygenic risk score for refractive error (R² = 6.04% vs. 6.01%). Our findings demonstrate that very few GWAS variants for refractive error show evidence of a departure from an additive mode of action and that accounting for non-additive risk variants offers little scope to improve the accuracy of polygenic risk scores for myopia.

全基因组关联研究（GWAS）表明，费希尔无穷小模型很好地描述了遗传对某些复杂疾病的影响，其中大量的多态性各自产生很小的影响。在费舍尔模型下，变异在基因座之间和基因座内都具有累加效应。然而，后一种假设与导致单基因疾病的显性或隐性稀有等位基因的常见观察结果不一致。在这里，我们寻找 GWAS 变异的非加性（显性或隐性）效应的证据，这些变异已知赋予高度遗传的数量性状——屈光不正的易感性。在 228,423 名个体的发现样本中检查了 146 个 GWAS 变异，这些个体的屈光不正表型是根据戴眼镜的年龄推断的，只有 8 人甚至具有非累加效应的名义证据 ( p < 0.05)。在接受直接屈光不正评估的 73,577 名个体的重复样本中，这 8 个变异中的 1 个具有非加性效应的强有力的独立证据（ ZMAT4内的 rs7829127， p = 4.76E−05），而另外 2 个具有暗示性证据（rs35337422）在RD3L中， p = 7.21E−03 和 rs12193446 在LAMA2中， p = 2.57E−02)。考虑非相加效应对屈光不正多基因风险评分的准确性影响极小（ R ² = 6.04% vs. 6.01%）。我们的研究结果表明，很少有屈光不正的 GWAS 变异显示出偏离加性作用模式的证据，并且考虑非加性风险变异几乎无法提高近视多基因风险评分的准确性。