Using in vivo genetic and physical interaction data for structure determination of protein assemblies

项目摘要 许多蛋白质通过形成大分子组装体来发挥功能。描述这些组件的结构 仍然是一个挑战。传统的结构生物学方法可以提供高- 分辨率原子结构，但通常需要纯化的样品，可能只描述少数构象。我们 建议使用体内遗传相互作用和定量交联质谱（qXL-MS）的数据 建立蛋白质组装结构模型的实验，使科学界能够解决 这些结构问题是目前传统结构生物学方法无法解决的。例如遗传 通过点突变上位性微阵列图谱（pE-MAP）平台和深度突变扫描进行相互作用作图 (DMS)已经成为强大的工具来询问蛋白质，在残留决议，在他们的背景下， 生物相关功能。类似地，体内qXL-MS方法非常适合于解剖物理组织。 蛋白质之间的相互作用，全方位的结构动力学，以及残基的构象变化 分辨率值得注意的是，体内遗传相互作用和交联实验可以在不同的温度下进行。 条件来确定蛋白质功能状态如何响应细胞环境的变化，这是一个问题。 很难用其他方法接近。然而，体内遗传相互作用和交联数据集通常是 噪声、稀疏和模糊，使得结构解释具有挑战性。为了充分发挥在 体内遗传和物理相互作用的数据，我们需要新的计算方法，最大限度地提高结构 从这些数据集中提取的信息。在这里，我们提出了一个全面的研究计划，以发展 构建稳定和瞬时蛋白质组装体的整合/混合结构模型的工具。我们将重点关注（1） 开发贝叶斯评分函数，其客观地量化体内的噪声和模糊性， 实验数据，从而提高模型的准确性和精度;（2）建立贝叶斯 分层模型来表示蛋白质组装的集合，因此允许应用程序 构象和组成异构系统;和（3）创建验证工具，以评估 使用体内数据获得的结构模型的精确度和准确度，因此允许明智地使用 模型最后，在与实验学家的密切合作下，我们将应用这些方法来确定 蛋白质组装体的结构已经折射到传统的结构生物学方法，包括 与HIV-1衣壳结合的牛痘病毒蛋白组装体、TRIM 5 α和与HIV-1衣壳结合的Ddis穿梭因子 蛋白酶体总之，我们将通过增加投入的种类来扩大结构生物学的范围 用于整合/混合结构建模的信息，从而允许生物结构建模 这些系统不适用于传统的结构生物学方法。我们的方法将在 开源集成建模平台（IMP）软件，并为全球蛋白质数据库做出贡献 （wwPDB）努力验证、存档和传播综合结构。

PROJECT SUMMARY Many proteins function by forming macromolecular assemblies. Describing the structures of these assemblies in their cellular environment remains challenging. Traditional structural biology approaches may provide high- resolution atomic structures but usually require purified samples and might describe only a few conformers. We propose using data from in vivo genetic interaction and quantitative cross-linking mass-spectrometry (qXL-MS) experiments to build structural models of protein assemblies, empowering the scientific community to address structural questions that are currently out of reach of traditional structural biology methods. For example, genetic interaction mapping by point-mutant epistatic miniarray profile (pE-MAP) platform and deep mutational scanning (DMS) have emerged as powerful tools to interrogate proteins, at a residue resolution, in the context of their biologically relevant functions. Similarly, in vivo qXL-MS approaches are well-suited to dissect physical interactions between proteins, a full range of structural dynamics, and conformational changes at residue resolution. Notably, in vivo genetic interaction and cross-linking experiments can be performed under varying conditions to determine how protein functional states respond to changes in the cellular environment, a problem difficult to approach by other methods. However, in vivo genetic interaction and cross-linking datasets are usually noisy, sparse, and ambiguous, making structural interpretation challenging. To fully realize the potential of in vivo genetic and physical interaction data, we need new computational methods that maximize the structural information extracted from these datasets. Here, we propose a comprehensive research program to develop tools to build integrative/hybrid structure models of stable and transient protein assemblies. We will focus on (1) developing Bayesian scoring functions that objectively quantify the noise and ambiguity in the in vivo experimental data, therefore increasing the accuracy and precision of the models; (2) building Bayesian hierarchical models to represent the ensembles of protein assemblies, therefore allowing the application to conformational and compositionally heterogeneous systems; and (3) creating validation tools to assess the precision and accuracy of structural models obtained using in vivo data, therefore allowing judicious use of the models. Finally, in close collaboration with experimentalists, we will apply these methods to determine the structures of protein assemblies that have been refractive to traditional structural biology methods, including vaccinia virus protein assemblies, TRIM5α bound to the HIV-1 capsid, and Ddis shuttling factors associated with the proteasome. In conclusion, we will expand the scope of structural biology by increasing the variety of input information used for integrative/hybrid structure modeling and thus allow structural modeling of biological systems that are not amenable to traditional structural biology approaches. Our methods will be implemented in the open-source Integrative Modeling Platform (IMP) software and contribute to the worldwide Protein Data Bank (wwPDB) effort to validate, archive, and disseminate integrative structures.