The selection and design of suitable synthetic paths are important issues that affect the economics and productivity of chemical processes including reactions and the discovery of new chemicals. However, exploration of reaction information is difficult even with reaction databases, causing path explosion that occurs due to the huge search space and conflicting constrains such as economics, safety, efficiency, etc. In this study, we propose an intelligent system ASICS (Advanced System for Intelligent Chemical Synthesis), which supports synthetic path design at the basic stages of research and process design, based on the hybrid generative exploration and exploitation of reaction knowledge base, encoding big data of patented reactions, and machine learning-based retrosynthetic prediction. Based on the pseudo A* search, ASICS generates optimal synthetic paths minimizing scores of the synthetic reaction value function, composed of the synthetic accessibility score, likelihood score and similarity score. The preference in searching between confirmed reaction spaces and unexplored reaction spaces through prediction can be selected by the user. The suggested hybrid approach, combining the reaction knowledge base with the retrosynthetic prediction model, generates feasible and low-cost synthetic paths beyond the accumulated information in patented reactions.

合适的合成路径的选择和设计是影响化学过程（包括反应和新化学品的发现）的经济性和生产率的重要问题。然而，即使使用反应数据库，反应信息的探索也很困难，由于巨大的搜索空间和经济、安全、效率等冲突约束而导致路径爆炸。在这项研究中，我们提出了一个智能系统 ASICS（高级系统）用于智能化学合成），支持在研究和工艺设计的基础阶段进行合成路径设计，基于反应知识库的混合生成探索和开发，编码专利反应的大数据，以及基于机器学习的逆合成预测。基于伪A*搜索，ASICS 生成最优合成路径，最小化合成反应价值函数的得分，由合成可访问性得分、似然得分和相似性得分组成。用户可以选择通过预测在已确认的反应空间和未探索的反应空间之间进行搜索的偏好。所建议的混合方法将反应知识库与逆合成预测模型相结合，可生成超出专利反应中积累信息的可行且低成本的合成路径。