A number of recent studies have investigated the role of de novo mutations in various neurodevelopmental and neuropsychiatric disorders. These studies attempt to implicate causal genes by looking for an excess load of de novo mutations within those genes. Current statistical methods for assessing this excess are based on the implicit assumption that all qualifying mutations in a gene contribute equally to disease. However, it is well established that different mutations can have radically different effects on the ultimate protein product and, as a result, on disease risk. Here, we propose a method, fitDNM, that incorporates functional information in a test of excess de novo mutational load. Specifically, we derive score statistics from a retrospective likelihood that incorporates the probability of a mutation being damaging to gene function. We show that, under the null, the resulting test statistic is distributed as a weighted sum of Poisson random variables and we implement a saddlepoint approximation of this distribution to obtain accurate p values. Using simulation, we have shown that our method outperforms current methods in terms of statistical power while maintaining validity. We have applied this approach to four de novo mutation datasets of neurodevelopmental and neuropsychiatric disorders: autism spectrum disorder, epileptic encephalopathy, schizophrenia, and severe intellectual disability. Our approach also implicates genes that have been implicated by existing methods. Furthermore, our approach provides strong statistical evidence supporting two potentially causal genes: SUV420H1 in autism spectrum disorder and TRIO in a combined analysis of the four neurodevelopmental and neuropsychiatric disorders investigated here.

中文翻译：

---

在多余的新突变负荷测试中纳入功能信息。

最近的许多研究调查了从头突变在各种神经发育和神经精神疾病中的作用。这些研究试图通过寻找因果基因中过量的从头突变来暗示这些基因。当前评估这种过量的统计方法是基于一个隐含的假设，即基因中所有合格的突变对疾病的贡献均相等。然而，众所周知的是，不同的突变对最终的蛋白质产物以及因此对疾病的风险具有根本不同的影响。在这里，我们提出了一种方法fitDNM，该方法将功能信息整合到了多余的从头突变负载测试中。具体而言，我们从回顾性可能性中得出分数统计信息，该回顾性可能性包括突变对基因功能造成损害的可能性。我们表明，在零值下，所得的测试统计量将作为Poisson随机变量的加权和分布，并且我们对该分布进行了鞍点近似以获得准确的p值。通过仿真，我们证明了在保持有效性的同时，我们的方法在统计功效方面优于当前方法。我们已经将该方法应用于神经发育和神经精神疾病的四个从头突变数据集：自闭症谱系障碍，癫痫性脑病，精神分裂症和严重智力障碍。我们的方法还涉及现有方法牵连的基因。此外，我们的方法提供了强有力的统计证据，支持两个潜在的因果基因：