Constraint Based Algorithms for Protein Folding

技术概述：材料研究部和数学科学部为该奖项提供资金，该奖项属于nsf范围内的数学科学优先领域，并为网络基础设施做出贡献。该奖项支持蛋白质折叠和反问题解决领域的理论和计算研究。PI计划将相位检索领域的成功解搜索策略应用于蛋白质折叠问题。该方法的关键要素是在欧几里得空间中的约束投影，以最小的努力将特定约束恢复到任意输入点。许多问题的解点都可以在两个约束集的交点上表述出来。对于这类问题，可以用相应的约束投影来定义动力系统，并将集合交问题的解编码在其不动点中。基于约束的算法是相位检索的首选方法，与主流采样算法相比，在蛋白质结构预测方面可能具有显著的优势。在相位检索中，过度确定的约束集赋予了显著的计算优势，当折叠序列设计良好时，期望获得类似的增益。用简单的异质聚合物模型进行的实验表明，这种方法可以扩展到现实模型，其中两个约束条件对应于链的几何形状和单体的堆积。该项目还将开发一种新的分布式计算形式，使基于约束的搜索的混沌动力学成为可能。下一代科学家和工程师将越来越依赖于共享的数据库和标准化的计算协议来进行他们的工作。这个项目的一个重要组成部分是开发这种工作环境的微型实现，称为“半蛋白质世界”。半蛋白质是具有高度简化性质的模型蛋白质，但它也面临着许多与真实蛋白质相同的挑战。通过一系列软件工具，包括基于网络的半蛋白质数据库，半专业研究人员可以通过网络访问设计和折叠半蛋白质，然后将他们的发现存入数据库。半蛋白质世界的设计将涉及伊萨卡地区的高中生和康奈尔材料研究中心NSF-REU项目的本科生的现场参与。非技术摘要：材料研究部和数学科学部为该奖项提供资金，该奖项属于nsf范围内的数学科学优先领域，并为网络基础设施做出贡献。该奖项支持蛋白质折叠领域的理论和计算研究。蛋白质是生物细胞的主要成分，在生物体的结构和功能中起着重要的作用。为了实现它们的生物学功能，蛋白质经历了一种自组装，以呈现特定的形状或折叠。折叠蛋白质的形状、折叠方式以及如何快速折叠是理解其功能的关键基本问题。现有的计算机模拟方法计算量非常大。在这里，PI将利用蛋白质折叠问题与显微镜之间的基本联系，寻求从不完美的数据（例如x射线衍射）中开发图像，以进一步开发一种强大的蛋白质折叠新算法，以克服长模拟时间的障碍。初步结果表明，该算法的计算效率将远远高于目前的方法。下一代科学家和工程师将越来越依赖于共享的数据库和标准化的计算协议来进行他们的工作。这个项目的一个重要组成部分是开发这种工作环境的微型实现，称为“半蛋白质世界”。半蛋白质是具有高度简化性质的模型蛋白质，但它也面临着许多与真实蛋白质相同的挑战。通过一系列软件工具，包括基于网络的半蛋白质数据库，半专业研究人员可以通过网络访问设计和折叠半蛋白质，然后将他们的发现存入数据库。半蛋白质世界的设计将涉及伊萨卡地区的高中生和康奈尔材料研究中心NSF-REU项目的本科生的现场参与。

TECHNICAL SUMMARY:The Division of Materials Research and the Division of Mathematical Sciences contribute funding to this award which falls under the NSF-wide Mathematical Sciences Priority Area and contributes to cyberinfrastructure. This award supports theoretical and computational research in the area of protein folding and the solution of inverse problems. The PI plans to apply a successful solution search strategy in the area of phase retrieval to the problem of protein folding. The key elements of this method are constraint projections in a Euclidean space that with minimal effort restore a particular constraint to an arbitrary input point. Many problems can be formulated in terms of their solution points being in the intersection of just two constraint sets. For such problems a dynamical system can be defined in terms of the corresponding constraint projections with the solution to the set intersection problem encoded in its fixed points. Constraint based algorithms are the method of choice in phase retrieval and may offer significant advantages, over mainstream sampling algorithms, in protein structure prediction.As in phase retrieval, where a significant computational advantage is conferred by overdetermined constraint sets, a similar gain is expected when folding sequences that are well designed. Experiments with simple heteropolymer models of proteins, where the two constraints correspond to chain geometry and monomer packing, show promise that this approach can be extended to realistic models. This project will also develop a novel form of distributed computing made possible by the chaotic dynamics of the constraint based search.The next generation of scientists and engineers will increasingly rely on shared data bases and standardized computing protocols in the conduct of their work. A significant component of this project is the development of a miniature realization of such a work environment called "semiprotein world". Semiproteins are model proteins with highly simplified properties, but which pose many of the same challenges posed by real proteins. Through a collection of software tools, including a web-based semiprotein data bank, semiprofessional researchers with web access will be able to design and fold semiproteins, and then deposit their findings in the data base. The design of semiprotein world will involve on-site participation of Ithaca area high school students and undergraduates in the Cornell Center for Materials Research NSF-REU program.NON-TECHNICAL SUMMARY:The Division of Materials Research and the Division of Mathematical Sciences contribute funding to this award which falls under the NSF-wide Mathematical Sciences Priority Area and contributes to cyberinfrastructure. This award supports theoretical and computational research in the area of protein folding. Proteins are major constituents of biological cells and play important roles in structure and function in living organisms. In order to carry out their biological function, the protein undergoes a kind of self-assemble to assume a particular shape or fold. The shape of a folded protein, how it folds and how it does it so quickly are key fundamental questions in understanding its function. Currently exiting computer simulation methods are very computationally intense. Here the PI will exploit fundamental connections between the problem of protein folding and microscopies that seek to develop an image from imperfect data, e.g. x-ray diffraction, to further develop a powerful new algorithm for protein folding to overcome the barrier of long simulation time. Preliminary results suggest that the algorithm will be far more computationally efficient than current methods.The next generation of scientists and engineers will increasingly rely on shared data bases and standardized computing protocols in the conduct of their work. A significant component of this project is the development of a miniature realization of such a work environment called "semiprotein world". Semiproteins are model proteins with highly simplified properties, but which pose many of the same challenges posed by real proteins. Through a collection of software tools, including a web-based semiprotein data bank, semiprofessional researchers with web access will be able to design and fold semiproteins, and then deposit their findings in the data base. The design of semiprotein world will involve on-site participation of Ithaca area high school students and undergraduates in the Cornell Center for Materials Research NSF-REU program.