SGER: Next Generation Computer-Assisted Thinking Tools for Plant Scientists.

当代实验生物学家可以接触到大量不同种类的关于许多不同生物体和生物系统的信息。虽然大部分信息仍在期刊上发表的论文中，但原始数据越来越多地驻留在可电子访问的数据库中。可用数据的数量和复杂性超过了任何单个研究人员的综合和推理能力，刺激了知识库的发展，知识库代表了比数据库更高抽象层次的生物系统信息。现有知识库有用性的一个严重限制是，关于系统的知识在存档时被有效地“冻结”了。与大部分现有证据相矛盾的观察结果通常在注释时被省略。然而，对生物系统如何运作的更深入的理解往往始于与现有知识相矛盾的观察。该项目的总体目标是创建计算工具，使实验生物学家能够探索有关生物系统的积累信息，而不排除访问相互矛盾的信息。该方法与目前用于构建知识库的方法明显不同，例如途径数据库Reactome （www.reactome.org）和BioCyc (www.biocyc.org)，它们存储基于专家输入和文献信息的当前接受的模型，并且必须随着知识的增长而修改和重写。这里采用的方法是将模型构建的成分存储在证据层面，并使生物学家能够通过制定和测试假设的熟悉设备持续积极地参与模型构建。假设本身用于查询存储的数据，方法是将每个假设分解为其组成关系，并提取为使假设有效而必须持有的所有显式和隐式断言。应用一组评估规则来测试这些断言是否与不同类型的数据一致，并向用户提供指向支持该假设的信息和数据的链接，以及指向与该假设相矛盾的信息和数据的链接。因此，不会将静态结论归档，而是存储所需的数据，以阐明系统中存在的关系。虽然实验者对人际关系的看法是根据已知的情况来检验的，但结论并不是强加的。相反，支持性信息和矛盾性都被报告，评估每一个的权重和重要性的任务留给实验者。这项工作有几个更广泛的影响。这种方法的重要性在于，它是实验生物学家的一种“思考”工具，而不是一种本质上是电子教科书的知识库。目前很少有计算机辅助思维工具可用，这个项目的成功可能会彻底改变实验生物学家的工作方式。这些工具将被整合到拟南芥信息资源（TAIR: www.arabidopsis.org）的运作中，这是一个社区数据库，在全球拥有14,000名固定用户。

Contemporary experimental biologists have access to a large body of disparate kinds of information about many different organisms and biological systems. Although much of the information is still in papers published in journals, primary data increasingly reside in electronically accessible databases. The volume and complexity of the available data exceed the synthetic and reasoning capacities of any individual researcher, stimulating the development of knowledge bases, which represent information about biological systems at a higher level of abstraction than do databases. A serious limitation on the usefulness of existing knowledge bases is that knowledge about the system is effectively "frozen" as it is archived. Observations that contradict the bulk of available evidence are generally omitted at the time of annotation. However, a deeper understanding of how biological systems operate often begins with an observation that contradicts existing knowledge. The overall objective of this project is to create computational tools that allow the experimental biologist to explore accumulated information about biological systems without precluding access to contradictory information. The approach differs markedly from those currently used to construct knowledge bases, such as the pathway databases Reactome (www.reactome.org) and BioCyc (www.biocyc.org), which store currently accepted models based on expert input and literature information and they must be revised and rewritten as the knowledge grows. The approach being taken here is to instead store the ingredients for model building at the evidence level and enable biologists to continuously and actively participate in model- building through the familiar device of formulating and testing hypotheses. The hypotheses themselves are used to query the stored data by breaking each hypothesis down into its constituent relationships and extracting all of the explicit and implicit assertions that must hold in order for the hypothesis to be valid. A set of evaluation rules is applied to test these assertions for agreement with different types of data and present the user with links to information and data that support the hypothesis, as well as links to those that contradict it. Thus, static conclusions are not archived, but instead the data required are stored to elucidate relationships that exist in the system. Although the experimenter's ideas about relationships are tested against what is known, conclusions are not imposed. Rather, both supporting information and contradictions are reported and the task of evaluating the weight and significance of each left to the experimenter. This work has several broader impacts. The importance of this approach is that it is a "thinking" tool for the experimental biologist, as opposed to a knowledge base, which is essentially an electronic textbook. Few computer assisted thinking tools are available at present and the success of this project could revolutionize how experimental biologists work. The tools will be integrated into the operation of The Arabidopsis Information Resource (TAIR: www.arabidopsis.org), a community database with 14,000 regular users world-wide.