Enhancing open data sharing for functional genomics experiments: Measures to quantify genomic information leakage and file formats for privacy preservation

项目摘要/摘要：随着大规模基因组数据的激增， 不同组学数据集的广度和深度以及在隐私主题中日益重要的 基因组数据科学中的个人。疾病和疾病的详细遗传和环境特征 条件依赖于功能基因组数据的大规模挖掘；因此，人们非常希望分享 数据尽可能广泛。然而，关注这类数据的隐私研究很少。一把钥匙 减少私人信息泄露的第一步是衡量 功能基因组数据，特别是不同数据文件类型的数据。为此，我们建议推导出 泛函中不同数据类型隐私信息泄露的信息论方法 基因组学数据。我们还将开发各种文件格式，以减少共享过程中的这种泄漏。我们会 在三个目标下进行隐私分析。首先，我们将开发可用于 量化从原始读取中泄露的敏感信息。我们将系统地分析链接攻击是如何 可以使用各种基因分型方法来实例化，例如单核苷酸变异和结构 来自原始读数、信号图谱、Hi-C相互作用矩阵和基因表达矩阵的变体调用。 其次，我们将研究不同的算法来实现对泛函的隐私保护变换 各种形式的基因组数据。特别是，我们将为原始序列创建隐私保护的文件格式 对齐地图、信号轨迹文件、三维交互矩阵和基因表达 包含来自多个个体的信息的量化矩阵。这将使我们能够研究 原始读取以外的敏感信息泄漏的来源，例如信号配置文件、拼接和 异构体转录和异常的三维基因组相互作用。第三，我们将调查 可以映射到原始人类功能基因组数据集中的微生物组的读数。我们将使用 推断的微生物信息来表征关于个人的私人信息，然后将 利用微生物信息从人类图谱中读取信息，以增加重新鉴定 第二个目标中描述的链接攻击的准确性。我们将使用这些工具来量化敏感的 来自大型测序项目的可用数据集中的信息和隐私保护文件格式， 例如ENCODE、癌症基因组图谱、1,000个基因组、gEUVADIS和基因组织 表达式项目。

Project Summary/Abstract: With the surge of large genomics data, there is an immense increase in the breadth and depth of different omics datasets and an increasing importance in the topic of privacy of individuals in genomic data science. Detailed genetic and environmental characterization of diseases and conditions relies on the large-scale mining of functional genomics data; hence, there is great desire to share data as broadly as possible. However, there is a scarcity of privacy studies focused on such data. A key first step in reducing private information leakage is to measure the amount of information leakage in functional genomics data, particularly in different data file types. To this end, we propose to to derive information-theoretic measures for private information leakage in different data types from functional genomics data. We will also develop various file formats to reduce this leakage during sharing. We will approach the privacy analysis under three aims. First, we will develop statistical metrics that can be used to quantify the sensitive information leakage from raw reads. We will systematically analyze how linking attacks can be instantiated using various genotyping methods such as single nucleotide variant and structural variant calling from raw reads, signal profiles, Hi-C interaction matrices, and gene expression matrices. Second, we will study different algorithms to implement privacy-preserving transformations to the functional genomics data in various forms. Particularly, we will create privacy-preserving file formats for raw sequence alignment maps, signal track files, three-dimensional interaction matrices, and gene expression quantification matrices that contain information from multiple individuals. This will allow us to study the sources of sensitive information leakages other than raw reads, for example signal profiles, splicing and isoform transcription, and abnormal three-dimensional genomic interactions. Third, we will investigate the reads that can be mapped to the microbiome in the raw human functional genomics datasets. We will use inferred microbial information to characterize private information about individuals, and then combine the microbial information with the information from human mapped reads to increase the re-identification accuracy in the linking attacks described in the second aim. We will use the tools to quantify the sensitive information and privacy-preserving file formats in the available datasets from large sequencing projects, such as the ENCODE, The Cancer Genome Atlas, 1,000 Genomes, gEUVADIS, and Genotype-Tissue Expression projects.