Predicting Feasible Organic Reaction Pathways Using Heuristically Aided Quantum Chemistry
Predicting Feasible Organic Reaction Pathways Using Heuristically Aided Quantum Chemistry
复制标题
DOI：
10.1021/acs.jctc.9b00126
复制
发表时间：
2019-07-01
影响因子：
5.5
通讯作者：
Aspuru-Guzik, Alan
中科院分区：
文献类型：
--
作者：
Rappoport, Dmitrij;Aspuru-Guzik, Alan
使用量子化学方法研究有机反应机理需要研究人员具有丰富的有机化学知识和第一性原理计算知识。之所以需要经验知识，是因为对有机化学反应势能面(PES)的任何合理完整的探索在计算上都是令人望而却步的。我们已经介绍了启发式辅助量子化学(HAQC)方法来模拟复杂的化学反应，该方法根据指导PES探索的化学启发式简单规则提取经验知识，并将其与使用量子化学方法进行结构优化相结合。HAQC方法使用启发式动力学标准来选择反应路径，这些路径不仅合理，即符合有机反应性的经验规则，而且在反应条件下也是可行的。在这项工作中，我们开发了启发式动力学可行性准则，该准则正确地预测了广泛范围的简单极性(取代、加成和消除)和周环有机反应(环化、σ向移动和环加成)的可行反应路径。与基于知识的反应机理预测方法相比，相同的动力学启发式方法成功地将反应路径分类为可行或不可行的反应机理集。我们讨论了HAQC的能量分布及其在化学反应的机器学习中的潜在应用。
Studying organic reaction mechanisms using quantum chemical methods requires from the researcher an extensive knowledge of both organic chemistry and first-principles computation. The need for empirical knowledge arises because any reasonably complete exploration of the potential energy surfaces (PES) of organic reactions is computationally prohibitive. We have previously introduced the heuristically-aided quantum chemistry (HAQC) approach to modeling complex chemical reactions, which abstracts the empirical knowledge in terms of chemical heuristics simple rules guiding the PES exploration and combines them with structure optimizations using quantum chemical methods. The HAQC approach makes use of heuristic kinetic criteria for selecting reaction paths that are not only plausible, that is, consistent with the empirical rules of organic reactivity, but also feasible under the reaction conditions. In this work, we develop heuristic kinetic feasibility criteria, which correctly predict feasible reaction pathways for a wide range of simple polar (substitutions, additions, and eliminations) and pericyclic organic reactions (cyclizations, sigmatropic shifts, and cycloadditions). In contrast to knowledge-based reaction mechanism prediction methods, the same kinetic heuristics are successful in classifying reaction pathways as feasible or infeasible across this diverse set of reaction mechanisms. We discuss the energy profiles of HAQC and their potential applications in machine learning of chemical reactivity.