AI-based genetic discovery for hearing loss

摘要 老年性听力损失是老年人最常见的疾病之一。影响听力的多种遗传因素 损失已经确定，但还有更多的损失有待确定；我们对 它们导致听力损失的机制是我们必须克服的障碍，如果我们要开发方法的话 用于预防(或逆转)与年龄相关的听力损失。没有一种模式生物的贡献超过 实验室老鼠对改善人类健康，以及老鼠模型塑造了我们对 哺乳动物的听觉系统。具有基因突变的小鼠已被用来识别对 听觉功能，以及描述导致听力损失的人类遗传因素。 在几个经过充分研究的近交系小鼠中，出现了一种具有寡基因基础的自发性听力损失 (A/J、DBA/2J、MA/MY、NOD/LTJ、NOR/LTJ、C57BR/CDJ、C57L/J)。我们最近开发的基于人工智能的 计算管道(GNNHap)确定了导致自发性听力损失的四个遗传因素 3株(A/J、DBA/2、NOD/LTJ)。然而，为了加快听力损失的基因发现步伐，这 Project将通过使其能够分析近交系基因组中存在的结构变异等位基因来增强我们的人工智能 菌株，并通过增加三个计算能力来确定候选基因的优先顺序。增强的人工智能将 能够：(I)确定已识别的小鼠候选基因的人类同源物中的等位基因是否 (Ii)分析表型数据库以确定一个小鼠品系 和(Iii)分析该基因中的基因表达数据 表达分析资源(GEAR)，以确定识别的候选小鼠基因(及其 人类同源基因)在耳朵中表达。然后将使用增强的计算工具来识别 4个品系(MA/MY、NOR/LTJ、C57BR/CDJ、C57L/J)听力损失的遗传因素。因为这是至关重要的 描述遗传效应机制，基因组工程用于产生敲击 (Ki)小鼠，其导致听力损失的遗传因素逆转为野生型。一个详细的 对这些Ki小鼠进行评估，以表征这些单独的(和联合的)效应 听力损失和耳蜗形态的突变。对它们的遗传效应机制的表征将 揭示一系列相互作用的寡聚因素是如何导致自发性听力损失的。作为一个延伸目标，我们 将使用其中的一些Ki小鼠来确定我们是否可以开发出一种新的噪音基因x环境模型- 诱发性听力损失。

Abstract Age-related hearing loss is one of the most common conditions in the elderly. Many genetic factors for hearing loss have been identified, but many more remain to be identified; and our lack of knowledge about the mechanisms by which they cause hearing loss is a barrier that must be overcome if we are to develop methods for preventing (or reversing) age-related hearing loss. No model organism has contributed more than the laboratory mouse to improving human health, and mouse models have shaped our understanding of the mammalian auditory system. Mice with genetic mutations have been used to identify genes that are critical for auditory function, and for characterizing human genetic factors that cause hearing loss. A spontaneous hearing loss with an oligogenic basis develops in several well-studied inbred mouse strains (A/J, DBA/2J, MA/My, NOD/LtJ, NOR/LtJ, C57BR/cdJ, C57L/J). Our recently developed AI-based computational pipeline (GNNHap) identified four causative genetic factors for spontaneous hearing loss in three strains (A/J, DBA/2, NOD/LtJ). However, to accelerate the pace of genetic discovery for hearing loss, this project will enhance our AI by enabling it to analyze structural variant alleles present in the genomes of inbred strains, and by adding three computational capabilities for prioritizing candidate genes. The enhanced AI will be able to: (i) determine if alleles within the human homologues of identified mouse candidate genes were associated with hearing loss in human GWAS; (ii) analyze a phenotypic database to determine if a mouse line with a knockout of a candidate gene has impaired hearing; and (iii) analyze gene expression data in the Gene Expression Analysis Resource (gEAR) to determine whether identified candidate murine genes (and their human homologues) are expressed in the ear. The enhanced computational tool will then be used to identify genetic factors for hearing loss in four strains (MA/My, NOR/LtJ, C57BR/cdJ, C57L/J). Since it is critical to characterize genetic effector mechanisms, state of the art genome engineering is used to generate knockin (KI) mice, which have a reversion of a causative genetic factor for hearing loss to wild type. A detailed evaluation of these KI mice is performed to characterize the individual (and combined) effect of these mutations on hearing loss and cochlear morphology. Characterization of their genetic effector mechanisms will reveal how a set of interacting oligogenic factors produce a spontaneous hearing loss. As a stretch goal, we will use some of these KI mice to determine if we can develop a novel gene x environment model for noise- induced hearing loss.