Technologies to Enable Privacy in Biomedical Databanks

描述（由申请人提供）：生物医学界正处于基因组学革命之中，有可能根据患者的基因组提供个性化医疗保健服务。为了利用最近的基因组学计划，科学家们已经开始研究发现个体基因组变异和临床表型之间的关系。大多数针对个人的记录的收集和分析都局限于特定的调查人员或机构；然而，科学家需要共享数据收集，以加强关联测试的统计能力，让其他人有机会验证他们的分析，并遵守政策要求。为了促进这一过程，世界各地的各种组织正在大力投资于数据库，以整合来自不同研究人员的患者特定记录。这种数据库能否广泛使用，取决于能否保护与共享记录相对应的个人的匿名性。尽管存在生物医学记录隐私的政策和技术方法，但它们不适合合并来自多个组织的记录的环境。特别是，各种调查表明，对个人特定生物医学记录的简单去识别使得集中的记录容易通过公共资源“重新识别”。我们研究的总体目标是基于正式的隐私和安全方法，为集中的个人特定生物医学记录开发一种新的数据保护模型。我们的解决方案将由一套技术组成，每一项技术都解决了生物医学数据库的构建和使用所面临的挑战。这些技术的发展将有三个具体目标：(1)建立一个工具来整合来自不同组织的研究参与者的生物医学记录，而不损害参与者的匿名性；(2)构建安全收集、存储和分析生物医学数据的方法，而不泄露个人记录；(3)检测和防止由于调查人员查询数据库而可能出现的政策违规行为。我们的方法将在软件中实现，使科学家和管理员不必处理不熟悉的隐私和安全协议的技术细节。最终的产品将是一种软件，使不同的数据持有者能够向中央生物医学数据库提交信息，科学家可以分析存储的记录，管理员可以监控系统的使用情况，以防止侵犯隐私。该软件将以模块化和可配置的方式设计，从而使用户能够选择最适合其环境的保护功能。为了证明我们的方法的适用性，本研究将专门解决现实世界的数据隐私挑战，这是多机构基因组全关联研究的瓶颈，但由此产生的模型和软件将可用于其他集中式数据库环境。我们相信，通过正式的隐私保护机制来管理生物医学记录，基于我们模型的数据库将能够以比目前更高的吞吐量支持研究。

DESCRIPTION (provided by applicant): The biomedical community is in the midst of a genomics revolution with the potential to personalize healthcare services to a patient's genome. To capitalize on recent genomics programs, scientists have initiated research to discover relationships between an individual's genomic variations and clinical phenotype. Most of the gathering and analysis of person-specific records has been localized to particular investigators or institutions; however, scientists need to share data collections to strengthen the statistical power of association tests, to allow others the opportunity to verify their analyses, and to comply with policy requirements. To facilitate this process, various organizations around the world are significantly investing in databanks to consolidate patient- specific records from disparate investigators. The availability of such databanks for wide-spread use is contingent on protecting the anonymity of the individuals that correspond to the shared records. Though policy and technical approaches for biomedical records privacy exist, they are inappropriate for environments that consolidate records from multiple organizations. In particular, various investigations demonstrate that the simple de-identification of person-specific biomedical records leave centralized records vulnerable to "re- identification" through public resources. The overarching goal of our research is to develop a novel data protection model for centralized person-specific biomedical records based on formal privacy and security methods. Our solution will be composed of a suite of technologies, each of which addresses a challenge for the construction, and use, of biomedical databanks. These technologies will be developed in three specific aims: (1) build a tool to integrate research participants' biomedical records from disparate organizations without compromising participants' anonymity, (2) construct methods to securely collect, store, and analyze biomedical data without revealing individual records, and (3) detect and prevent policy violations that can arise as a consequence of investigators queries to the databank. Our methods will be implemented in software that shields scientists and administrators from handling the technical details of unfamiliar privacy and security protocols. The final product will be software that enables disparate data holders to submit information to a centralized biomedical databank, scientists to analyze the stored records, and administrators to monitor the use of system for privacy violations. The software will be designed in a modular and configurable manner, thus enabling users to pick and choose which protection features are most appropriate to their environment. To demonstrate the applicability of our methodology, this research will specifically address a real world data privacy challenge that is a bottleneck for multi-institutional genome wide association studies, but the resulting models and software will be reusable for other centralized databanking environments. We believe that by managing biomedical records through formal privacy protection mechanisms, databanks based on our model will be able to support research with greater throughput than the current status quo.