With rapid advances in high-throughput sequencing technology, millions of single-nucleotide variants (SNVs) can be simultaneously genotyped in a sequencing study. These SNVs residing in functional genomic regions such as exons may play a crucial role in biological process of the body. In particular, non-synonymous SNVs are closely related to the protein sequence and its function, which are important in understanding the biological mechanism of sequence evolution. Although statistically challenging, models incorporating such SNV annotation information can improve the estimation of genetic effects, and multiple responses may further strengthen the signals of these variants on the assessment of disease risk. In this work, we develop a new weighted empirical Bayes method to integrate SNV annotation information in a multi-trait design. The performance of this proposed model is evaluated in simulation as well as a real sequencing data; thus, the proposed method shows improved prediction accuracy compared to other approaches.

中文翻译：

---

使用多个特征的加权经验贝叶斯风险预测模型。

随着高通量测序技术的飞速发展，可以在测序研究中同时对数百万个单核苷酸变体（SNV）进行基因分型。这些位于外显子等功能基因组区域的SNV可能在机体的生物过程中发挥关键作用。特别是，非同义SNV与蛋白质序列及其功能密切相关，这对于理解序列进化的生物学机制非常重要。尽管统计上具有挑战性，但包含此类SNV注释信息的模型可以改善遗传效应的估计，并且多重响应可以进一步增强这些变异对疾病风险评估的信号。在这项工作中，我们开发了一种新的加权经验贝叶斯方法，将SNV注释信息集成到多特征设计中。该模型的性能在仿真和实际测序数据中得到了评估。因此，与其他方法相比，该方法显示出更高的预测精度。