基于晶体生长机理模型的制药过程结晶溶剂设计方法研究

Solution crystallization is an important and efficient purification/separation unit operation in active pharmaceutical ingredient (API) production processes. The requirements of crystal products include not only purity and yield, but also crystal morphology (habit). To obtain optimal crystal products, it is crucial to reveal the influence of solvents and mechanisms for the control of crystal growth. Nowadays, experimental trial and error is still the most widely used crystallization solvent design method. Therefore, it is urgent to develop a systematic crystallization solvent design method from the interdisciplinary studies of supramolecular chemistry, condensed matter physics, and product design principles. In this project, the organic small molecules gliflozin drugs are studied as an example for the crystallization solvent design. First, the step-wise regression method based on decision tree is used to establish the solvent evaluation index system to quantitatively regulate crystal morphology, based on the mechanism of the effects of crystal surface – solvent interaction energy ES on the growth rates of each crystal surface, to reveal the nature of solvation effect to the crystal morphology using modified attachment energy model. Second, the group contribution data is regressed by using the group contribution and machine learning methods for the solvent evaluation index system. Then, the computer-aided molecular design model, which integrates the established mechanism model, is developed for the design of crystallization solvent. Mathematical optimization model is established for the design of crystallization solvent to consider the requirements of crystal morphology, yield, green and safety simultaneously. Finally, experiments are implemented for verification. The results of the project are able to be expanded to various APIs for the efficient and clean production of pharmaceutical processes, and contribute to the pharmaceutical products upgrade and crystal engineering.

溶液结晶是原料药生产过程中重要的高效纯化、分离单元。晶体产品对纯度、产率及晶体形貌（晶习）均有严格要求。揭示溶剂影响机制并控制晶体生长内在规律是获得理想结晶产品的关键。目前溶剂选择方法多为实验试错筛选，亟需从超分子化学、凝聚态物理、产品设计原理交叉探寻结晶溶剂设计方法。本项目以列净类药物为例，首先利用基于决策树的逐步回归法，定量构建对晶面—溶剂吸附能ES影响显著且变量互不相关的溶剂评价指标体系，基于修正附着能模型揭示溶剂对晶体形貌的影响规律，建立晶体形貌定量调控的结晶溶剂机理模型；进一步，通过耦合机理模型与计算机辅助分子设计方法，利用基团贡献与机器学习方法对评价指标进行基团数据回归，建立数学规划模型并进行优化求解，开展同步集成晶体形貌、产率、绿色安全的结晶溶剂设计方法；最后进行实验验证。项目成果可拓展应用至多种原料药，并为制药过程高效清洁生产奠定基础，为医药产品升级及晶体工程发展做出贡献。

溶液结晶是原料药（API）生产过程中重要的分离单元操作。对晶体产品的要求不仅包括纯度和收率，还包括晶体形貌。溶剂是影响晶体形貌的重要因素之一。因此，揭示溶剂对晶体形貌的影响机制对于控制结晶过程具有重要意义。然而，如何选择/设计合适的结晶溶剂以满足晶体形貌、收率和绿色安全等方面的要求，仍然是结晶领域亟待解决的问题之一。本项目提出了一种基于晶体形貌定量调控模型的结晶溶剂设计框架。首先，利用分子动力学（MD）来预测溶剂中的晶体形貌。然后通过逐步回归方法选择合适的溶剂描述符，建立晶体长径比和溶剂描述符间的定量关系。以苯甲酸和布洛芬为例来验证所建立的定量调控模型的准确性和外延性。进一步，将晶体形貌定量调控模型与计算机辅助分子设计方法（CAMD）进行耦合。结晶溶剂设计问题可以表示为混合整数非线性规划模型（MINLP），并采用分步求解策略进行求解。最后，将结晶溶剂设计框架应用于苯甲酸和布洛芬两个案例，并进行实验验证。

内容获取失败，请点击重试