Advanced approaches to protein structure prediction

摘要 在过去的四十年里，基因组测序的成功导致了基因组之间的差距迅速扩大。 已知蛋白质序列的数量以及已知蛋白质结构和功能的数量。因为 蛋白质序列本身不能告诉我们每个分子在细胞中的作用， 结构和功能信息严重阻碍了当代生物学和医学研究的进展。 这些理解上的差距强烈要求高效的计算方法，以实现自动化，但高度 准确的蛋白质结构预测和功能注释。私家侦探的实验室在 开发和传播高质量的结构生物信息学方法，这些方法已被广泛使用， 全球社会。在这个项目中，实验室寻求开发新的先进方法， 蛋白质四级结构预测基于PI实验室以前开发的工具和数据库， 新的基于深度神经网络的技术将扩展到剩余水平的链内和链间接触， 距离图预测这些预测将被用来约束构象搜索空间 基于线程的片段组装模拟，目的是显着提高准确性， 单体蛋白质和蛋白质-蛋白质相互作用（PPI）结构建模的成功率，特别是对于 在蛋白质数据库中缺乏同源模板的困难目标。接下来，PPI网络的结构 信息将用于帮助阐明蛋白质的多层次生物学和生物医学功能 分子，包括突变引起的蛋白质稳定性变化和人类疾病预测。很长的- 该项目的长期目标是显著提高蛋白质结构预测的技术水平， 缩小蛋白质序列信息丰富与蛋白质结构缺乏之间的差距， 功能数据，从而显着提高结构生物信息学的有用性和影响。胜任这 项目还将有助于揭示控制序列之间基本关系的一般原则， 蛋白质分子的结构和功能。

Abstract The success of genome sequencing over the last four decades has resulted in a rapidly increasing gap between the number of known protein sequences and the number of known protein structures and functions. Because protein sequence on its own cannot tell us what each molecule does in cells, the large-scale absence of protein structure and function information severely hinders the progress of contemporary biological and medical studies. These gaps in understanding strongly call for efficient computational approaches for automated, yet highly accurate protein structure prediction and function annotation. The PI’s lab has a successful track record in developing and disseminating high-quality structural bioinformatics methods which have been widely used by the global community. In this project, the lab seeks to develop new advanced methods for both tertiary and quaternary protein structure prediction. Built on the tools and databases previously developed in the PI’s lab, new deep neural-network based techniques will be extended to residue-level intra- and inter-chain contact- and distance-map predictions. These predictions will then be used to constrain the conformational searching space of threading-based fragment assembly simulations, with the aim to significantly improve the accuracy and success rate of structure modeling of monomeric proteins and protein-protein interactions (PPIs), especially for the difficult targets that lack homologous templates in the Protein Data Bank. Next, the structure and PPI network information will be used to help elucidate multiple levels of biological and biomedical functions for protein molecules, including mutation-induced changes in protein stability and human disease predictions. The long- term goals of this project are to significantly improve the state of the art of protein structure prediction and to narrow the gap between the abundance of protein sequence information and the dearth of protein structure and function data, thus significantly enhancing the usefulness and impact of structural bioinformatics. Success in this project will also help reveal the general principles governing the fundamental relations across sequence, structure and function of protein molecules.