Developing Computational Tools for Predicting and Designing Function-Enhancing Enzyme Variants

项目摘要 高效突变酶的先验预测和设计被广泛认为是“圣杯”。 因为它将使研究人员找到有效的酶变体来降解环境- 精神污染物，进行精细化学品的后期功能化，治疗疾病。定向进化 已被广泛应用于寻找化学反应的最佳酶变体，但如何加速 由于序列、结构和序列之间的未知关系，筛选周期仍然是一个关键的障碍 酶催化动力学。为了克服这一挑战，PI一直在开发三种计算- 框架结构：1)集成的酶结构-功能数据库，IntEnzyDB，提供干净和表格 用于数据驱动建模的数据；2)用于评估酶反应物种的新软件模块RosettaQM 使用量子力学方法的交互；以及3)高通量工作流EnzyHTP，它允许COM- 酶变异体的推定筛选。在这份Mira提案中，PI强调，通过这些工具， 为预测和设计新的酶催化剂提供新的计算工具。首先，PI将制定一项 利用多态动力学评分函数预测突变对表观酶动力学的影响 IntEnzyDB中存储的酶学数据和基于QM的酶反应物种相互作用评分 RosettaQM(项目-1)。PI计划开发一个动态计分功能，以说明多个- 沿着反应路径的三重反应状态，与现有的计算理性工程不同 强调一个假设的过渡状态的稳定的战略。多状态动态得分将 通过实验验证来预测FR29酯酶的效率提高突变以进行概念验证 以及用于环境污染物降解的氟乙酸酯脱卤酶FAcD。第二，PI将制定一项 集成的酶预测协议，增强了分子模拟和机器学习模型，以 设计酶变体以适应非天然底物，以实现类药物分子的后期功能化- Eules(项目2)。将进一步开发EnzyHTP以融入机器学习模型，以实现多 目的预测有益突变，评价突变对酶静电环境的影响。 底物定位，底物-酶相互作用，以及酶的稳定性、溶解性和混杂。恩- 将在设计新的基团转移酶以适应S-腺苷时对zyHTP进行实验检验 蛋氨酸类似物对治疗广泛性骨质疏松症的药物依维诺米星的后期功能化作用 耐抗生素细菌感染的谱系。总而言之，拟议的研究将带来 用于生物医学和生物医学的功能增强酶变体的虚拟预测和设计的计算工具 生物催化用途。

Project Summary The a priori prediction and design of efficient mutant enzymes are broadly recognized as a “Holy Grail” in chemistry and biology because it will allow researchers to find effective enzyme variants to degrade environ- mental pollutants, conduct late-stage functionalization of fine chemicals, and treat diseases. Directed evolution has been widely applied to identify optimal enzyme variants for chemical reactions, but how to accelerate the screening cycles remains a critical roadblock due to the unknown relationship between sequence, structure, and kinetics for enzyme catalysis. To overcome this challenge, the PI has been developing three computational in- frastructures: 1) an integrated enzyme structure-function database, IntEnzyDB, that provides clean and tabulated data for data-driven modeling; 2) a new software module, RosettaQM, for evaluating enzyme-reacting species interactions using quantum mechanical methods; and 3) a high-throughput workflow, EnzyHTP, that allows com- putational screening of enzyme variants. Enabled by these tools, in this MIRA proposal, the PI emphasizes advancing new computational tools to predict and design new enzyme catalysts. First, the PI will develop a multistate kinetic scoring function to predict the influence of mutation on apparent enzyme kinetics by leveraging the enzymology data stored in IntEnzyDB and the QM-based enzyme-reacting species interaction scoring in RosettaQM (Project-1). The PI plans to develop a kinetic scoring function that accounts for contributions of mul- tiple reactive states along a reaction pathway and is distinct from existing computational rational engineering strategies that emphasize the stabilization of one hypothetical transition state. The multistate kinetic scoring will be experimentally validated to predict efficiency-enhancing mutations for FR29 esterase for a proof of concept and fluoroacetate dehalogenase FAcD for environmental pollutant degradation. Second, the PI will develop an integrated enzyme predicting protocol that augments molecular simulations and machine-learning models to design enzyme variants to accommodate non-native substrates for late-stage functionalization of drug-like mol- ecules (Project-2). EnzyHTP will be further developed to incorporate machine learning models to achieve multi- objective prediction of beneficial mutations, evaluating the impact of mutations on enzyme electrostatic environ- ment, substrate positioning, substrate-enzyme interactions, and enzyme stability, solubility, and promiscuity. En- zyHTP will be experimentally examined in the design of new group-transferases to accommodate S-adenosyl methionine analogues for late-stage functionalization of everninomicin, a promising drug for treating a broad spectrum of antibiotic-resistant bacterial infections. In summary, the proposed research will deliver enabling computational tools for virtual prediction and design of function-enhancing enzyme variants for biomedical and biocatalytic uses.