Combining Voice and Genetic Information to Detect Heterogeneity in Major Depressive Disorder

项目总结 这项应用旨在通过结合遗传基因来促进我们对严重抑郁障碍(MDD)的理解 并分析MDD患者的语音模式，以识别亚型。MDD是主要原因 残疾是全世界最常见的疾病之一，然而，相对于其他常见疾病，人们对其起源知之甚少。 有效的治疗方法较少，花在试图了解它是如何发生以及如何治疗上的费用要少得多 治愈它。目前的治疗方法相对无效，高达50%的患者难以治愈，许多患者遭受痛苦 严重复发。了解MDD背后的机制已被公认为是一项重大的任务 全球心理健康面临的挑战。因此，开发MDD的新治疗方法是公共卫生的主要优先事项。 MDD研究的一个主要挑战是异质性的存在。存在多个子类型的 MDD已经被怀疑了很长一段时间，可能会混淆适当治疗这种疾病的能力 用现有的治疗方法，以及很难确定MDD的原因作为发展的前奏 新的治疗方法。然而，找到亚型一直很难。考虑到人们说话的方式可以反映 情绪的变化，我们希望声音能够预测情绪，因此有可能被用作 认识到异质性。初步数据显示，结合遗传数据的高维发声 从记录中提取的特征可用于识别子类型。此外，遗传数据的使用 允许我们将语音特征归入到没有录音的大型生物库中，从而使探索成为可能 发声特征与丰富的临床重要指标之间的关系。我们探索了这种力量 对MDD作出诊断，预测病情严重程度和其他临床特征。将我们的方法应用于 将为临床管理提供信息，改进诊断，改进治疗，并帮助开发新的 治疗法

PROJECT SUMMARY This application aims to advance our understanding of major depressive disorder (MDD) by combining genetic information and analyzing speech patterns of those with MDD to identify subtypes. MDD is the leading cause of disability throughout the world, yet, relative to other common disorders, less is known about its origins. There are less effective treatments and much less is spent on trying to understand how it arises and how to cure it. Current treatments are relatively ineffective, with up 50% of patients refractory and many suffering severe recurrence. Understanding the mechanisms underlying MDD has been recognized as a grand challenge in global mental health. Thus, developing new treatments for MDD is a major priority for public health. A major challenge for MDD research is the presence of heterogeneity. The existence of multiple subtypes of MDD has been suspected for a long time, and likely confounds the ability to treat the disorder appropriately with existing treatments, as well as making it hard to identify the causes of MDD as a prelude to developing new treatments. However finding subtypes has been hard. Given that the way people talk can reflect alterations in mood, we expect voice to be able to predict mood, and hence potentially be used as biomarker to recognize heterogeneity. In preliminary data show that in combination with genetic data high-dimensional vocal features extracted from recordings can be used to identify subtypes. Furthermore, the use of genetic data allows us to impute voice features into large biobanks where no recordings exist, making it possible to explore the relationship between vocal features and a rich array of clinically important indicators. We explore the power of voice to make a diagnosis of MDD, to predict severity and other clinical features. Applying our approach to will inform clinical management, improving diagnosis, refine treatment and aid the development of new treatments