STORAGE AND ANALYSIS OF FLOW CYTOMETRY DATA AT THE CELLULAR LEVEL USING RELATIO

这个子项目是许多研究子项目中利用 资源由NIH/NCRR资助的中心拨款提供。子项目和 调查员(PI)可能从NIH的另一个来源获得了主要资金， 并因此可以在其他清晰的条目中表示。列出的机构是 该中心不一定是调查人员的机构。 使用关系数据库技术和超级计算资源在细胞级别存储和分析流式细胞仪数据。主要研究人员：基思·博伊斯生物信息学免疫耐受网络副主任/加州大学旧金山分校2585自由港路-套房207匹兹堡，宾夕法尼亚州15238 V：412 820 8807(X2312)F：412-820-8912电子邮件：kbayce@munetolerance.org简介：免疫耐受网络的基本目标是加速人类疾病新耐受疗法的临床开发。ITN在许多疾病领域进行临床试验和耐受性分析研究：胰岛移植、肾脏和肝脏移植、自身免疫性疾病、过敏症和哮喘。ITN使用的主要机械检测技术之一是流式细胞术。ITN数据中心开发了一种存储和检索流式细胞术实验数据(面板、试管、使用的抗体等)的新方法。并在总结/样本层面上得出结果。在11个临床试验中，有超过2500个样本，有近100万行来自最初的流式细胞仪实验室评估的结果。原始结果以其原生的行业标准FCS文件格式存档在文件存储库中。对原始数据的任何重新分析都需要从FCS文件加载原始蜂窝/事件数据。虽然FCS文件是按照批准的标准设计的，但如果没有FCS分析软件包或了解FCS数据格式/标准的数据适配器，则无法对其进行分析。在与许多其他数据中心和流式细胞仪实验室讨论后，这种存储和重新分析的方法似乎是一种常见的方法。摘要：我们提议与网络基础设施伙伴关系和TeraGrid就初始开发分配(DAC)进行合作，以创建一个蜂窝事件数据的Oracle数据库。一旦创建了数据库，我们希望探索与TeraGrid在数据挖掘、分析、可视化和基于TeraGrid功能的模拟方法方面的潜在合作，这将导致未来的一个中型或大型分配项目。作为数据加载和存储的一部分，我们建议将蜂窝级事件数据保留在关系数据库中，而不是关联的FCS文件中。虽然可以在仪器运行期间设置事件的数量，但典型的ITN分析在每个分析的流式细胞仪试管中从20,000到100,000个细胞/事件不等。ITN目前有超过2700个分析样本，涉及11个不同的临床试验。这可能会导致需要存储45,000,000个单个细胞事件的结果，每个事件通常具有37个荧光强度测量。ITN目前每个项目有4到5个衡量标准。在100亿个事件之间，事件的数量很可能每年都会增加。在正在进行的试验中，大部分分析集中于对来自试验多个治疗分支的参与者的派生化验数据的时程分析。随着临床试验的完成，将会有更多的实例需要对免疫耐受数据进行交叉试验分析。这些分析可能需要对大量的原始细胞事件进行分析。此外，在多个试验中进行数据挖掘和探索潜在的免疫耐受生物标志物可能需要探索所有这些细胞数据，以及与临床结果、表型、基因类型、人口统计学、治疗/治疗以及ITN进行的其他机制分析的结果相关的许多其他数据维度。问题解决：流式细胞术是丰富的细胞信息来源，可用于广泛的生物信息学应用。我们相信，通过使用TeraGrid处理和关系数据库技术，这些应用程序和想法中的许多将被启用。这种DAC可以证明，可以显著减少当前与在FCS文件中存储细胞事件数据相关的方法施加的与原始流式细胞仪数据分析相关的处理、可伸缩性和系统限制。目前对FCS文件的处理通常需要在内存中管理许多数据加载和操作以及汇总/过滤功能，而数据库管理系统可以更好地处理这些功能。这将使生物信息学专业人员能够集中精力构建高性能的并行分析和建模功能，而不是文件、数据和内存管理。

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Storage and analysis of flow cytometry data at the cellular level using relational database technology and supercomputing resources. Principal Investigator: Keith Boyce Associate Director Bioinformatics Immune Tolerance Network / UCSF 2585 Freeport Road - Suite 207 Pittsburgh, PA 15238 V: 412 820 8807 (x 2312) F: 412-820-8912 Email: kboyce@immunetolerance.org Introduction: The fundamental objective of the Immune Tolerance Network, or ITN, is to accelerate the clinical development of new tolerance therapeutics in human disease. The ITN conducts clinical trials and tolerance assay studies in a number of disease areas: Islet Transplantation Kidney & Liver Transplantation Autoimmune Diseases Allergy & Asthma. One of the primary mechanistic assay technologies used by the ITN is flow cytometry. The ITN Data Center has developed a novel method of storing and retrieving flow cytometry experimental data (panels, tubes, antibody used etc.) and derived results at the summary / specimen level. With over 2500 specimens across 11 Clinical Trials there are nearly a million rows of derived results from the original flow cytometry laboratory assessment. Raw results are archived in a file repository in their native, industry standard FCS file format. Any re-analysis of the raw data requires a loading of the raw cellular / event data from the FCS files. While FCS files are designed to an approved standard, they are not analyzable without an FCS analysis software package or data adapters that are aware of the FCS data format / standard. After discussions with many other Data Centers and Flow Cytometry laboratories this storage and re-analysis approach appears to be a common method. Abstract: We are proposing to work with the CyberInfrastructure Partnership and TeraGrid on an initial Development Allocation (DAC) to create an Oracle database of cellular event data. Once the database is created we would like to explore potential collaborations with TeraGrid on data mining, analysis, visualization, and simulation methods based on TeraGrid capabilities which would lead to a Medium or Large Allocation project in the future. As part of our data loading and storage we are proposing to retain the cellular level event data in a relational database rather than in an associated FCS file. While the number of events can be set during the running of the instrument, typical ITN assays vary from 20,000 to 100,000 cells / events per flow cytometry tube analyzed. The ITN currently has over 2700 analyzed specimen across 11 different clinical trials. This would potentially result in the need to store results for 45,000,000,000 individual cellular events with typically 3 7 fluorescent intensity measurements per event. The ITN currently has 4 or 5 measures per event. The numbers of events will most likely increase on an annual basis between 10 20 billion events. During on-going trials much of the analysis focuses on time-course analysis of derived assay data from participants across multiple treatment arms of the trial. As the clinical trials complete there will be more instances where cross-trial analysis of immune tolerance data will need to be performed. These analyses could require the analysis of a very large number of these raw cellular events. Additionally, data mining and exploration for potential biomarkers of immune tolerance across multiple the trials will likely need to explore all of this cellular data plus many other data dimensions related to clinical outcomes, phenotypes, genotypes, demographics, treatments / therapies and results from other mechanistic assays conducted by the ITN. Problem addressed: Flow cytometry is a rich source of cellular information which can be used in a wide variety of bioinformatics applications. We believe that by using TeraGrid processing and relational database technology many of these applications and ideas will be enabled. This DAC can demonstrate that processing, scalability, and system limitations related to the analysis of raw flow cytometry data which are imposed by the current methodologies related to storing cellular event data in FCS files can be significantly reduced. Current processing of FCS files often requires in-memory management of many data loading and manipulation and summarization / filtering functions which can be better handled by database management systems. This would enable bioinformatics professionals to focus their efforts on building high performance parallel analysis and modeling functions rather than file, data and memory management.