Postdoctoral Research Fellowships in Biology for FY 2009

本行动资助NSF 2009年度博士后研究奖学金。该奖学金支持Firas Khatib的一项名为“自动化人类直觉以提高蛋白质结构预测”的研究和培训计划。这项研究的主办机构是华盛顿大学，赞助科学家是大卫·贝克博士。这个项目利用人类和计算机的联合力量来建立与疾病相关的蛋白质的精确模型。确定蛋白质的正确结构有助于研究人员成功地用药物靶向它们。求解蛋白质的实验方法既耗时又昂贵，这使得计算解决蛋白质折叠问题成为必要。即使有了当前的超级计算和分布式计算能力，也需要新的算法来解决构象采样这一主要的计算瓶颈之一。因为一个蛋白质链可以有很多不同的构象，有效地探索搜索空间是至关重要的。解决这个问题的一种新方法是利用人类解决谜题的直觉和模式识别技能，结合运行Rosetta（大卫·贝克研究小组开发的预测算法）的计算机。Foldit是一个基于rosetta的交互式程序，允许世界各地的用户在他们的家用电脑上实时直接操作蛋白质结构模型。Foldit的最终目标是利用计算机程序的结果，通过利用人类自然的3D问题解决能力来改进结构预测算法。该项目决定了游戏玩家可以改进结构预测问题的哪些方面，并将这些新技术应用到Rosetta的自动化协议中。人类直觉的成功自动化将是计算科学的一项非凡成就。培训目标包括发展编程和计算技能以及教学和指导能力。Foldit用户能够与世界各地的其他人分享他们的解决方案，利用不同人的优势，以改善全球结果。在当地的科学教育博物馆和当地的高中，Foldit正在向孩子们和家长们做报告，希望Foldit可以在世界各地的化学课程中使用，创造下一代蛋白质文件夹，然后反过来，新的结构预测算法可以基于这些新专家的蛋白质折叠技能开发出来。这为全世界的参与者提供了蛋白质折叠知识的即时和相关应用，希望能加强理解基础科学的动力。此外，由于世界上任何人都可以为Foldit做出贡献，因此这项研究建立了一个全球协作网络。开发的算法和方法将被公布，Rosetta软件的源代码将免费和公开地提供给科学界。

This action funds an NSF Postdoctoral Research Fellowship for FY 2009. The fellowship supports a research and training plan entitled "Automating human intuition to improve protein structure prediction" for Firas Khatib. The host institution for this research is the University of Washington, and the sponsoring scientist is Dr. David Baker.This project uses the combined power of humans and computers to build accurate models of disease-related proteins. Determining the correct structure of a protein helps researchers successfully target them with drugs. Experimental methods for solving proteins are time consuming and expensive, making a computational solution to the protein-folding problem necessary. Even with the current supercomputing and distributed computing power available, new algorithms are needed to solve one of the main computational bottlenecks, conformational sampling. Because a protein chain can have so many different possible conformations, efficient exploration of the search space is crucial. A new approach to this problem is the use of human puzzle-solving intuition and pattern recognition skills combined with computers running Rosetta, the prediction algorithm developed in David Baker's research group. Foldit is an interactive Rosetta-based program allowing users around the world to directly manipulate protein structure models on their home computers in real-time. The ultimate goal of Foldit is to use the results from the computer program to improve structure prediction algorithms by capitalizing on natural human 3D problem-solving skills. The project determines what aspect of the structure prediction problem game-playing humans can improve and implement these new techniques into Rosetta's automated protocol. The successful automation of human intuition would be a remarkable achievement in computational science. Training objectives include development of programming and computational skills as well as teaching and mentoring abilities. Foldit users are able to share their solutions with others around the world, taking advantage of the strengths that different people have, in order to improve results globally. Presentations on Foldit are being given to children and parents at the local science and educational museum and at local high schools in the hope that Foldit could be used in chemistry courses around the world, creating the next generation of protein folders, and then in turn new structure prediction algorithms could be developed based on these new experts' protein-folding skills. This gives worldwide participants an immediate and relevant application for protein-folding knowledge, hopefully reinforcing the motivation to understand the underlying science. In addition, because anyone in the world can contribute to Foldit, this research establishes a global collaborative network. The algorithms and methods developed will be published, and the Rosetta software source code will be freely and openly available to the scientific community.