CompBio: A New Paradigm of Protein Threading: simultaneous backbone threading and side-chain packing prediction.

了解蛋白质的三级结构对于理解蛋白质的生物学功能和作用机制是必不可少的。由于测序基因组数量的快速增长，而实验确定的蛋白质结构数量的增长速度相对较慢，因此计算蛋白质结构的重要性正在增加。在这里，我们建议开发一种新的范例，利用侧链信息进行基于线索的蛋白质结构预测。虽然在过去的十年里，基于残基的方法使得线程技术在计算上预测蛋白质结构是可行的，但我们清楚地看到，现在迫切需要更准确的预测技术，能够提供更高分辨率的结构数据，以满足快速增长的结构和功能基因组学研究的需求。为了解决这个具有挑战性的问题，我们将开发一个新的计算框架来解决一个通用的线程问题，在这个框架中，主干线程和侧链填充被同时预测。具体地说，我们将(A)开发一个新的能量函数，该函数将残基水平的信息用于主链线索和原子水平的信息用于侧链填充；(B)开发一个新的算法框架来解决广义线索问题；(C)扩展该算法框架以处理更一般的线索问题，如约束线索问题；(D)测试和评估新的线索能量函数和算法，以展示利用侧链信息进行蛋白质结构预测的可行性和威力；以及(E)开发一个免费访问的基于网络的预测服务器，使整个研究社区受益。据我们所知，这个项目可能是第一次系统地努力推广当前的线程范例，以包括线程过程中的详细侧链信息。虽然侧链信息应该有助于显著提高蛋白质结构的预测精度和分辨率，但广义穿线问题提出了一些非常具有挑战性的计算问题。我们预计，这种新的线程能力，当完全开发和实施时，将显著提高蛋白质结构预测的技术水平。此外，本项目中开发的新的线程能量函数和算法技术将被证明对其他研究人员开发蛋白质结构预测能力有用。我们相信，我们的新线索框架将引领新一代蛋白质结构预测技术的发展，它不仅将提供比现有线索方法更准确的主干结构，而且还将提供当前线索技术遗漏的很大一部分结构信息，即侧链。该项目为本科生和研究生提供了一个理想的培训场所，让他们学习生物信息学工具开发，以解决复杂的生物学问题。基于该项目的研究成果，将开设一门新的课程《蛋白质结构预测与建模的算法》，面向本科生和研究生。

The knowledge of the tertiary structure is essential to understanding of the biological function and functional mechanism of a protein. The importance of computational solution to protein structures is increasing owing to the rapid growth in the number of sequenced genomes, and the relatively slow growth rate in the number of experimentally determined protein structures. Here we propose to develop a new paradigm for threading-based protein structure prediction using side-chain information. While residue-based approaches have made threading computationally feasible to predict protein structures in the past decade, we clearly see an urgent need now for more accurate prediction techniques that can provide structural data at higher resolution, to meet the rapidly growing need of structural and functional genomics studies. To address this challenging issue, we will develop a novel computational framework for solving a generalized threading problem, in which backbone threading and side-chain packing are predicted simultaneously. Specifically, we will (a) develop a new energy function that combines residue-level information for backbone threading and atom-level information for side-chain packing; (b) develop a novel algorithmic framework for solving the generalized threading problem; (c) extend this algorithmic framework to deal with more general threading problems such as constrained threading problems; (d) test and evaluate the new threading energy functions and algorithms to demonstrate the feasibility and the power of using side-chain information in protein structure prediction, and (e) develop a freely accessible web-based prediction server to benefit the entire research community. To the best of our knowledge, this project probably represents the first systematic effort in generalizing the current threading paradigm to include the detailed side-chain information during the threading process. While the side-chain information should help to significantly improve the prediction accuracy and resolution of protein structures, the generalized threading problem raises some very challenging computational problems. We expect that this new threading capability, when fully developed and implemented, will significantly improve the state of the art of protein structure prediction. In addition, the new threading energy functions and the algorithmic techniques developed in this project will prove to be useful to other researchers in their own development of protein structure prediction capabilities. We believe that our new threading framework will lead the way in developing a new generation of prediction techniques for protein structures, which will not only provide much more accurate backbone structures compared to the existing threading methods but also provide a big portion of the missing structural information by the current threading techniques, i.e., the side-chains. This project provides an ideal training ground for both undergraduate and graduate students to learn bioinformatics tool development for solving complex biological problems. A new course on "Algorithms for Protein Structure Prediction and Modeling" will be developed based on the research results of this project, which will be offered to both undergraduate and graduate students.