Planning and Workflow Management for a Virtual Laboratory for Structural Biology

我们建议建立一个基础设施，即虚拟实验室，以支持结构生物学知识的共享，规划实验，并促进严格的科学发现过程。我们称它为虚拟实验室，因为它将允许个人坐在连接到互联网的计算机前控制实验，跟踪过去由一个成熟的团队进行的实验的最相关方面，或者在关键时刻获得建议。结构生物学和其他自然科学如物理或化学的发现过程相当复杂，它涉及一些乏味和耗时的活动，知识共享，结果分析，回溯。虚拟实验室将允许结构生物学家减少获得结果的时间，提高结果的质量，并通过允许科学家专注于实验和发现过程来重组人类的参与。为了减少获得结果的时间，我们需要高效的组件、完全集成的实验、模拟和建模环境，以及某种形式的知识共享和管理。为了提高解的质量，我们需要在反馈回路中充分集成传感器，并使用最优的模型参数。最后但并非最不重要的一点是，我们需要自动化简单的任务，如数据迁移、格式转换和为每个计算任务选择最佳运行条件。在这一领域获得高质量结果所需的高度复杂程度激励着我们的努力。基础设施的一些组件特定于结构生物学，例如用于构建知识库的接口，但基本机制如规划和工作流执行引擎是通用的，可以用于具有类似特征的其他类别的问题。在这项拨款申请中，我们建议使用低温电子显微镜方法开发用于3D结构确定的代理，并设计一个基于计划引擎和知识管理系统的基于PETRI网模型的集成工作流执行引擎。我们的所有程序都是在我们的网站http://bond.cs.purdue.edu.的开源许可下分发的该项目将有助于教育和支持结构生物学实验室的知识共享。

We propose to create an infrastructure, a Virtual Laboratory, to support sharing of knowledge in structural biology, to plan an experiment, and to facilitate a rigorous process of scientific discovery. We call it a virtual laboratory because it will allow an individual sitting in front of a computer connected to the Internet, to control an experiment, to follow the most relevant aspects of an experiment conducted in the past by a well established group, or to get advice at a critical junction. The process of discovery in structural biology, and other natural sciences like physics or chemistry is fairly complex, it involves some tedious and time consuming activities, sharing of knowledge, analysis of the results, backtracking. The virtual laboratory willallow structural biologists to reduce the time to obtain the results, improve the quality of the results, and restructure the human involvement by allowing a scientist to concentrate on the experiment and the discovery process. To reduce the time to obtain results we need efficient components, fully integrated experiment, simulation, and modeling environments, and some form of knowledge sharing and management. To improve the quality of the solution we need to fully integrate sensors in the feedback loop and to use optimal model parameters. Last, but not least, we need to automate simple tasks like data migration, format conversion, and selection of optimal running condition for each computational task. The high level of sophistication necessary to obtain quality results in this field motivates our effort.Some components of the infrastructure are specific to structural biology e.g. the interfaces for building the knowledge base, but the basic mechanisms e.g., the planning and the workflow enactment engines, are general and can be used for other classes of problems with similar characteristics. In this grant application we propose to develop agents for 3D structure determination using cryo TEM methods and to design an integrated workflow enactment engine based upon a Petri Net model with a planning engine and a knowledge management system.All our programs are distributed under open source licenses from our Web site: http://bond.cs.purdue.edu. The project will contribute to education and support knowledge sharing in structural biology laboratories.