An integrated toolkit combining computational systems biology techniques with molecular dynamics simulations to delineate functionality of GPCRs

项目总结/摘要 在过去的十年里，G蛋白偶联的高分辨率结构的解析出现了爆炸式的增长。 受体（GPCR）及其与几种G蛋白的复合物，统称为转导蛋白。GPCRs 是动态蛋白质，并且以多种功能构象状态存在。三维的比较 GPCR的非活性和活性状态的结构已经导致了残基对距离的鉴定， 激活后发生明显变化。这种残基对被称为"激活微动开关"。分子动力学 (MD)模拟是一个有吸引力的工具，用于识别（i）对GPCR激活至关重要的残基对，（ii） 参与从配体结合位点到G蛋白偶联位点的变构通讯的残基对，和（iii） GPCR：G蛋白界面中的残基对有助于它们的偶联强度和选择性。 虽然我们生成长时间尺度动力学轨迹的能力呈指数级增长，但 MD模拟已经在很大程度上使用GPCR的先验知识进行了分析。迫切需要掌握 采用无偏的、数据驱动的系统生物学工具对长时间尺度的MD轨迹数据进行分析挖掘 关于残基运动的知识提供了GPCR中变构通信网络的信息。我们 这项资助的首要目标是应用贝叶斯网络（BN）建模，这是一种可解释的机器学习方法。 方法，MD模拟轨迹数据的GPCR：G蛋白复合物，以确定残留物中 各种GPCR结构区域有助于配体选择性。以网络为中心的方法还没有被 使用，到目前为止，分析高维残差对MD模拟数据。特别是BN建模，具有吸引人的 属性（可解释性，数据表示的概率性质，统计验证，拓扑工具 比较分析），目前部署的二次MD模拟数据分析方法，如主 残基对的成分分析（PCA），缺乏。 我们建议使用BN建模与大规模MD轨迹（多个短和多个长轨迹） 非活性状态GPCR和完全活性状态GPCR：G蛋白复合物的分析，以（i）识别激活微动开关， 活化后显示大规模构象变化的残基对，以及（ii）概述残基网络 参与从激动剂结合位点到G蛋白偶联位点的变构通讯。我们也（三） 鉴定GPCR：G蛋白界面中有助于选择性偶联至特定G蛋白家族的残基 蛋白质（Gs、Gi和Gq）。在目标2中，我们将（iv）使用BN序列和动态BN（DBN）来剖析时间相关事件 模型来描绘和审查导致大规模过渡的残留网络。 重要的是，交付成果将包括一个前所未有的"工具包"（算法+软件）， 系统生物学工具，用于分析可推广到任何蛋白质复合物的MD模拟轨迹数据。

Project Summary / Abstract In the last decade there has been an explosion of resolved high-resolution structures of G-protein coupled receptors (GPCRs) and their complexes with several G-proteins, collectively known as transducer proteins. GPCRs are dynamic proteins, and exist in multiple functional conformational states. Comparisons of three-dimensional structures of the inactive and active states of GPCRs have led to identification of residue pair distances that show distinct changes upon activation. Such residue pairs are known as “activation microswitches”. Molecular Dynamics (MD) simulations is an attractive tool for identifying (i) the residue pairs that are critical to GPCR activation, (ii) residue pairs involved in allosteric communication from the ligand binding site to the G protein coupling site, and (iii) residue pairs in the GPCR:G protein interfaces that contribute to their coupling strength and selectivity. While our ability to generate long time scale dynamics trajectories have increased exponentially, the results of MD simulations have largely been analyzed using prior knowledge of the GPCRs. There is a critical need for adopting the unbiased, data-driven, systems biology tools to analyze long time scale MD trajectories data to mine knowledge on the residue motions that provide information on allosteric communication network in GPCRs. Our overarching goal in this grant is to apply Bayesian Network (BN) modeling, an interpretable machine learning methodology, to the MD simulation trajectories data on GPCR:G protein complexes in order to identify the residues in various GPCR structural regions that contribute to ligand selectivity. Network-centered approaches have not been used, so far, to analyze high-dimensional residue pairs MD simulation data. BN modeling, in particular, has attractive properties (interpretability, probabilistic nature of the data representation, statistical validation, tools for topology comparison and analysis) that presently deployed secondary MD simulation data analysis methods, such as principal component analysis (PCA) of residue pairs, lack. We propose to use BN modeling with large scale MD trajectories (multiple short and multiple long trajectories) of inactive state GPCRs and fully active state GPCR:G protein complexes to (i) identify the activation microswitches, the residue pairs that show large scale conformational changes upon activation, and (ii) outline the residue network involved in the allosteric communication from the agonist binding site to the G-protein coupling sites. We will also (iii) identify the residues in the GPCR:G protein interface that contribute to selectivity in coupling to specific family of G proteins (Gs, Gi and Gq). In aim 2, we will (iv) dissect time-correlated events using BN series and dynamic BN (DBN) models to delineate and scrutinize the residue networks that lead to large scale transitions. Importantly, the deliverables will include an unprecedented “toolkit” (algorithms + software) incorporating system biology tools for analyzing MD simulation trajectories data that are generalizable to any protein complexes.