Interactive Medline Search Engine Integrating External Information

描述(申请人提供)：高通量实验方法极大地加速了生物医学研究。微阵列分析、全基因组关联研究、深度测序和脑成像等方法减少了数据生成和收集的瓶颈。然而，理解高通量数据的生物学意义是一个重大挑战1。正如Bota和Swanson所指出的，现在“记忆、评估和合成神经科学文献远远超出了个体研究人员的能力范围，无论多么出色，即使是在网络结构、生理学或化学等受限领域也是如此”。我们认为，问题的一个关键部分是在现有文献和数据的背景下，对基于高通量实验数据绘制高维函数关系的支持不足。流行的搜索解决方案，如PubMed/Google Scholar，主要是为了高效地检索最相关的信息而设计的，而不是为了探索性假设的发展。这些解决方案缺乏我们提议的系统将提供的几个关键功能，即通过旨在开发假设的文献和数据库探索来理解高通量数据的生物学所需的功能： 概述在熟悉的生物环境中的Medline搜索结果，以促进探索： 以反映检索到的记录的内在生物关系的图形概述来呈现搜索结果将比仅仅是潜在相关记录的线性列表更有效。这样的综述最好是来自多个生物学背景，也应该支持对属性数据和模式关联的有效交互探索，以从多个角度得出不明显的关系。 对不同算法、生物实体和数据源的查询支持：一个检索算法不能适应所有情况。Medline查询需要支持诸如基因ID和基因组位置等生物实体。Medline数据库需要得到外部数据源的补充，如本体、路径和各种数据库，这些数据库包含从实验数据得出的经过精选的信息。 第三方插件和跨应用功能集成的开放架构：需要第三方数据和功能插件的支持，以增强解决方案的功能和适配。开放式架构将允许使用来自其他解决方案的中间数据和/或功能。 结合这些功能，我们建议开发一个名为PubViz的系统，该系统将通过集成探索与高通量实验结果相关的文献和数据，更有效地支持神经生物学家对主要精神障碍潜在分子机制的假设。我们还将进行系统的需求评估和用户测试，以确保我们开发的功能有效地满足用户的需求。在我们现有的组件功能原型的基础上，PubViz将提供一个在帮助科学家制定假设方面超过其他系统的查询和分析环境。它将把Medline搜索结果与来自外部资源的数据和信息相结合，并在有用和可用的多种生物背景下以可视和交互的方式确定关系的位置。创建这些组合创新和人机界面(HCI)设计绝非易事，但考虑到我们在可视化Medline探索解决方案开发、数据分析和集成以及可用性和有用性研究方面的试点工作和经验，创建这些组合创新和人机界面设计是可行的。此外，将这个项目的重点放在神经生物学和精神障碍上，这是一个我们拥有丰富经验的研究领域，将有助于我们更有效地满足关键用户需求和功能。此外，我们开发的解决方案应该适用于其他生物医学研究领域。

DESCRIPTION (provided by applicant): High throughput experimental methods have accelerated biomedical research dramatically. Approaches such as microarray analysis, genome-wide association studies (GWAS), deep sequencing and brain imaging reduce bottlenecks in data generation and collection. Understanding the biological significance of high throughput data, however, is a major challenge 1. As pointed out by Bota and Swanson, it is now "far beyond the grasp of individual investigators, no matter how brilliant, to remember, evaluate, and synthesize the neuroscience literature, even in restricted domains like network structure, physiology, or chemistry" 2. We argue that a key part of the problem is insufficient support for drawing high dimensional functional relationships based on high throughput experimental data in the context of existing literature and data. Prevailing search solutions, such as PubMed/Google Scholar, are mainly designed for retrieving the most relevant information efficiently but not for explorative hypothesis development. These solutions lack several key functionalities that our proposed system will provide, functionalities required for understanding the biology of high throughput data through literature and database explorations that aim at hypothesis development: Overview of Medline search results in familiar biological contexts to facilitate exploration: Presenting the search results in graphic overviews reflecting inherent biological relationships of the retrieved records will be more effective than a linear list of potentially relevant records alone. Such overviews, ideally from multiple biological contexts, should also support efficient interactive exploration of attribute data and pattern associations for deriving non-obvious relationships from multiple perspectives. Query support for different algorithms, biological entities and data sources: One retrieval algorithm will not fit all situations. Biological entities such as gene IDs and genomic locations need to be supported for Medline queries. The Medline database needs to be supplemented by external data sources such as ontology, pathway, and various databases containing curated information derived from experimental data. Open architecture for third party plug-ins and cross-application function integration: The support of third party data and function plug-ins are needed to enhance the functionality and the adaptation of a solution. Open architecture will enable the use of intermediate data and/or functions from other solutions. Incorporating these functions, we propose to develop a system called PubViz that will more effectively support neurobiologists' needs for developing hypotheses on molecular mechanisms underlying major mental disorders through integrated exploration of literature and data related to high throughput experimental results. We will also conduct systematic needs assessments and user tests to ensure that functions we develop match users' needs effectively. Building on our existing component function prototypes, PubViz will provide a query and analysis environment that exceeds other systems in helping scientists work toward formulating hypotheses. It will integrate Medline search results with data and information from external resources and situate relationships visually and interactively in multiple biological contexts that are useful and usable. Creating these combined innovations and human-computer interface (HCI) designs is non-trivial but is feasible given our pilot work and experience in visual Medline exploration solution development, data analysis and integration and usability and usefulness studies. Additionally, focusing this project on neurobiology and mental disorders, a research domain in which we have extensive experience will help us address critical user needs and functionalities more effectively. Moreover, the solution we develop should be adaptable to other biomedical research domains.