D3SC: Discovery and Optimization of Chiral Catalysts Guided by Chemoinformatics

有了这个奖项，美国国家科学基金会化学部的化学催化项目支持了斯科特·E。伊利诺伊大学香槟分校化学系的丹麦。丹麦教授正在开发一种新的催化剂开发模式，将“多样性导向的综合”的创造力与信息学的计算能力结合起来。在数据驱动的化学发现科学（D3 SC）计划下，该计划的基础是发明和实施高分辨率描述符，这些描述符能够以计算机可读的形式准确反映分子的化学性质。现有技术不足以提供所需的粒度水平，以确保计算机生成的假想催化剂库的性质准确反映分子的真实特征。然而，最重要的是，开发一个真正通用的、计算指导的工作流程来优化分子功能，可以通过优化功能材料（传感器、LED、粘合剂、储能系统等）的性能来影响整个化学性质谱。发现和优化工业相关化学过程的催化剂。该提案的目标总结在一个发现导向计划的基本组成部分，该计划将计算分析与实验相结合，即：（1）基于给定的支架计算机生成假设催化剂结构的大量库，然后计算每个库成员的描述符，（2）多样性分析以产生代表性的“训练集”，（3）训练集的合成，（4）给定反应中训练集的评价，（5）将经验输出与分子性质相关联的数学模型的开发和验证，（6）将该模型应用于催化剂的虚拟库，（7）最佳预测催化剂的合成和评价，和（8）重复步骤4-8直到获得所需的输出。三维定量结构选择性关系（3D-QSSR）建模不仅通过识别高性能催化剂结构，而且通过提供新的框架来阐明控制催化剂对任何给定化学转化的活性和对映选择性的结构特征，从而影响不对称催化，存在巨大的潜力。最重要的是，这个框架应该普遍适用于任何类型的分子性质或功能的优化。这些活动是理想的研究生和博士后同事的智力和实践培训。理论与实验的相互作用是科学方法的本质。学生提出了假设的结果计划实验，他们必须学会收集和解释数据，以证实或消除假设。他们也成为专家合成化学家，精通科学程序员，并在计算化学，数据科学和机器学习能力。这项活动的统一主题是开发一个计算指导的工作流程，从而创建可用于优化无数化学反应的普遍适用的分子训练集。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。

With this award, the Chemical Catalysis Program of the NSF Division of Chemistry is supporting the research of Professor Scott E. Denmark of the Department of Chemistry at the University of Illinois at Urbana-Champaign. Professor Denmark is developing a new paradigm for the development of catalysts that combines the creative power of "diversity-oriented synthesis" with the computational power of informatics. Under the data-driven discovery science in Chemistry (D3SC) initiative, the foundation of this program is the invention and implementation of high-resolution descriptors that are able to accurately reflect the chemical properties of molecules in a form that is readable by a computer. The current state of the art is insufficient to provide the level of granularity needed to assure that the properties of computer generated libraries of hypothetical catalysts are accurate reflections of the true character of the molecules. Most importantly, however, the development of a truly general, computationally guided workflow for the optimization of molecular function can impact the entire spectrum of chemical properties from optimizing the performance of functional materials (sensors, LEDs, adhesives, energy storage systems, etc.) to the discovery and optimization of catalysts for industrially relevant chemical processes.The goals of this proposal are summarized in the basic components of a discovery oriented program that combines computational analysis with experimentation, namely: (1) in-silico generation of a massive library of hypothetical catalyst structures based on a given scaffold followed by calculation of descriptors of each library member, (2) diversity analysis to generate representative "training set", (3) synthesis of training set, (4) evaluation of the training set in a given reaction, (5) development and validation of a mathematical model that correlates empirical output with molecular properties (6) application of that model to the virtual library of catalysts, (7) synthesis and evaluation of best predicted catalysts, and (8) repeat steps 4-8 until desired output is achieved. Great potential exists for 3D-Quantitative Structure Selectivity Relationship (3D-QSSR) modeling to impact asymmetric catalysis, not only by identifying high performance catalyst structures, but also by providing a new framework to elucidate the structural features that govern the activity and enantioselectivity of catalysts for any given chemical transformation. Most importantly, this framework should be universally applicable to the optimization of any kind of molecular property or function. These activities are ideal for the intellectual and practical training of graduate students and postdoctoral coworkers. The interplay of theory and experiment is the essence of the scientific method. Students are presented with hypotheses for the outcome of planned experiments and they must learn to collect and interpret data to substantiate or eliminate the hypothesis. They also become expert synthetic chemists, proficient scientific programmers, and competent in computational chemistry, data-science, and machine learning. The unifying theme of this activity is the development of a computationally guided workflow that leads to the creation of universally applicable training sets of molecules that can be used to optimize myriad chemical reactions.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.