ClinTAD: A Tool for Improving Clinical CNV Interpretation

项目摘要/摘要 Dna微阵列是一种常规的儿科临床诊断试验，用于鉴定基因组拷贝数变异(Cnv)。 这可能是自闭症、发育迟缓和多种先天性畸形的原因。这项测试也是 经常用于产前环境中，以找出超声波发现的胎儿异常的基因组原因，或 在出生后预测潜在的表型。目前对CNV的临床解释指南仅侧重于 CNV断裂点中包含的基因的特征。然而，最近对染色质的研究 利用Hi-C或相关技术的体系结构已经证明，CNV也可以破坏结构 拓扑相关结构域(TADS)。TAD是物理DNA相互作用的“邻居”，服务于 几种功能，包括防止异位基因-增强子相互作用。这种轻微的干扰可能会导致 与CNV区域外的基因转录的病理性变化有关，这些变化最终导致了 疾病。这一建议的中心假设是只关注临床上CNV区域内的基因 解释，关键的基因组信息在DNA微阵列解释中最终被完全忽视 导致患者漏诊。为了解决这个问题，我们最近开发了免费使用的 软件ClinTAD(www.clintad.com；J Hum Genet(2019))，帮助在服用时解释CNV 考虑到潜在的TAD干扰。据我们所知，这是同类软件中第一个尝试 将TADS应用于临床CNV判读。虽然ClinTAD v1.0目前作为决策支持提供 作为协助临床实践的工具，它目前在易用性和预测能力方面都受到限制。 进一步增强ClinTAD的实用性激发了我们在此提议的两个目标：1)我们的目标是优化 ClinTAD作为临床决策支持和研究工具，允许纳入TAD边界 从不同的数据集，启用用于大型案例队列分析的API，添加解释工具 在SNP阵列上发现纯合子区域，并允许创建一个未识别的数据库，其中 用户可以根据TAD中断上传怀疑致病的病例。2)我们的目标是改善 ClinTAD通过机器学习识别最具预测性的致病特征的预测能力 在一个大型的可公开获得的CNV队列中，以及通过结合最近描述的卷积神经 基于网络的模型，可以根据CNV断点预测TAD中断。在这份提案中，我们 目的使ClinTAD成为在TAD干扰的背景下解释CNV的首要工具。我们的长- 学期目标是建立一个用户协作网络，使我们能够识别最多的患者 TAD干扰引起的临床表型的概率。这样一个独特的患者队列可以 形成了第一个评估Hi-C作为临床测试的实用性的同类试验的基础。

PROJECT SUMMARY/ABSTRACT DNA microarray is a routine clinical pediatric diagnostic test to identify genomic copy number variants (CNVs) that could be causative of autism, developmental delay, and multiple congenital anomalies. This test is also used frequently in the prenatal setting to find genomic causes of fetal anomalies found by ultrasound, or to predict potential phenotypes postnatally. Current clinical guidelines for interpretation of CNVs focus solely on the characteristics of genes contained within the CNV breakpoints. However, recent studies on chromatin architecture, utilizing Hi-C or related techniques, have demonstrated that CNVs can also disrupt the structure of topologically associated domains (TADs). TADs are “neighborhoods” of physical DNA interactions that serve several functions, including the prevention of ectopic gene-enhancer interactions. This TAD disruption can lead to pathogenic alterations in transcription of genes outside the CNV region that are ultimately causative of disease. The central hypothesis of this proposal is by only focusing on genes within the CNV region for clinical interpretation, critical genomic information is being entirely ignored in DNA microarray interpretation, ultimately leading to missed diagnoses for patients. To address this issue, we have recently developed the free-to-use software ClinTAD (www.clintad.com; J Hum Genet (2019)) to assist in the interpretation of CNVs while taking potential TAD disruption into account. To our knowledge, this is the first software of its kind to attempt to integrate TADs into clinical CNV interpretation. While ClinTAD v1.0 is currently available as a decision support tool to assist in clinical practice, it is currently limited both in its ease-of-use as well as its predictive power. Further enhancing the utility of ClinTAD motivates the two Aims of our proposal here: 1) We aim to optimize ClinTAD as both a clinical decision support and research tool by allowing incorporation of TAD boundaries from different datasets, enabling an API for analysis of large case cohorts, adding interpretation tools for Regions of Homozygosity found on SNP array, and allowing for creation of a de-identified database where users can upload cases with suspicion for pathogenicity based on TAD disruption. 2) We aim to improve the predictive power of ClinTAD through machine learning to identify the most predictive features of pathogenicity in a large, publicly available CNV cohorts, as well as by incorporating a recently-described convolutional neural network-based model which can predict TAD disruption as a function of CNV breakpoints. In this proposal we aim to make ClinTAD the premier tool for the interpretation of CNVs in the context of TAD disruption. Our long- term goal is to build a collaborative network of users that will enable us to identify patients with the most probability of having clinical phenotypes caused by TAD disruption. Such a unique patient cohort could then form the basis of a first-of-its kind trial to evaluate the utility of Hi-C as a clinical test.