MFB: Novel Graph Neural Networks to Understand, Predict, and Design Allosteric Transcription Factors

在这个生物技术分子基础（MFB）项目中，佐治亚理工学院的Corey J. Wilson教授、Matthew J. Realff教授和Yao Xie教授正在利用新的实验和机器学习策略来理解、预测和设计一个被称为转录因子的蛋白质家族的变抗通讯，该家族调节生命系统中的基因表达。蛋白质变构是一种重要的蛋白质功能，它使功能蛋白的不同部分之间的交流成为可能。由于缺乏对变构机制的了解，科学家和工程师无法设计出这种至关重要的功能。通过结合分子生物学和人工智能的工具，该项目旨在在分子水平上，特别是在单个氨基酸水平上，破译自然发生和工程转录因子的结构/活性模式。原则上，理解控制变构通讯的规则将使研究人员能够设计出新的转录因子，用于各种高影响力的应用，如操纵肠道细菌的组成。该项目涉及生物化学、生物物理学、工程学和机器学习方法的融合，将促进学生跨越传统学科界限的参与。除了多样化的学生参与外，该项目更广泛的影响将包括在机器学习和生物工程领域开发创新的教学模块。该项目将有助于培养多样化和敬业的STEM（科学、技术、工程和数学）劳动力，为终身为研究、教育及其整合做出贡献奠定坚实的基础。蛋白质变构是一种至关重要的蛋白质功能，在分子水平上理解它是一个令人烦恼的问题。该项目的目标是破译变构信号穿越支架的潜在分子机制，通过几种自然发生和工程转录因子，从更广泛的LacI/GalR蛋白同源物家族中替代变构控制。一般来说，变构通讯涉及非相邻氨基酸位置的网络；因此，传统的两两计算方法（例如，分子力学模拟和相关的计算机辅助蛋白质设计策略）在理解和设计先验变构网络方面的作用有限。因此，本项目寻求开发新的机器学习方法和互补的实验策略，以加速科学进步，改变研究和设计变构沟通的本质。这个项目有可能导致关于单一折叠变构网络的起源和构建（设计）的范式转变。此外，该项目中开发的机器学习方法原则上可以应用于指定蛋白质系统之外的其他复杂网络问题，例如蒸馏塔序列、通信系统和电网系统。该项目由化学部、化学、生物工程、环境和运输系统部以及信息和智能系统部联合支持。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

In this Molecular Foundations for Biotechnology (MFB) project, Professors Corey J. Wilson, Matthew J. Realff, and Yao Xie at the Georgia Institute of Technology are leveraging both novel experimental and machine learning strategies to understand, predict, and design allosteric communication in a family of proteins called transcription factors that regulate gene expression in living systems. Protein allostery is an important protein function which enables communication between different parts of a functional protein that are widely separated. Our lack of understanding of the mechanism of allostery prevents scientist and engineers from designing this critically important function. By combining the tools of molecular biology and artificial intelligence, this project aims to decipher structure/activity patterns for naturally occurring and engineered transcription factors at the molecular level, specifically at the level of individual amino acids. Understanding the rules that govern allosteric communication would, in principle, enable investigators to design new transcription factors for a variety of high-impact applications such as manipulating the composition of bacteria in the gut. This project involves a blend of biochemistry, biophysics, engineering, and machine learning approaches that will facilitate student engagement across traditional disciplinary boundaries. In addition to diverse student involvement, the broader impacts of this project will include the development of innovative pedagogical modules in the areas of machine learning and biological engineering. This project will contribute to the development of a diverse and engaged STEM (science, technology, engineering and mathematics) workforce, building a firm foundation for a lifetime of contributions to research, education, and their integration. Protein allostery is a vitally important protein function that has proven to be a vexing problem to understand at the molecular level. The goal of this project is to decipher the underlying molecular mechanisms by which the allosteric signal traverses the scaffold across several naturally occurring and engineer transcription factors with alternate allosteric controls from the broader LacI/GalR family of protein homologues. In general, allosteric communication involves networks of non-neighboring amino acid positions; therefore, traditional pairwise computational approaches (e.g., molecular mechanics simulations, and related computer-aided protein design strategies) are of limited use in understanding and designing allosteric networks a priori. Accordingly, this project seeks to develop novel machine learning approaches and complementary experimental strategies to accelerate scientific progress and transform the nature of studying and designing allosteric communication. This project has the potential to lead to a paradigm shift with regard to the origins and construction (design) of allosteric networks in a single fold. Moreover, the machine learning approaches developed in this project can in principle be applied to other complex network problems beyond the designated protein systems – e.g., distillation column sequences, communication systems, and power grid systems.This project is jointly supported by the Division of Chemistry, the Division of Chemical, Bioengineering, Environmental and Transport Systems, and the Division of Information and Intelligent Systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.