目的 基于化学成分群加和性分子描述符建立中药浸膏粉溶化性预测模型,并进行验证。方法 采用原位浊度传感器评价中药浸膏粉溶化性,对测试方法条件进行优化并进行验证;以13种代表性中药(广藿香、当归、薄荷、瞿麦、银柴胡、乌梅、葛根、细辛、益母草、苦参、熟地黄、荆芥和川芎)浸膏粉为研究对象,检索每种中药所含化学成分,根据成分的分子描述符计算出每种中药化学成分群平均性质,进而采用偏最小二乘(partial least squares,PLS)法建立该平均性质与对应中药溶液浊度值的关联预测模型并验证。结果 中药浸膏粉溶化性客观评价方法的重复性和耐用性良好;对从13种中药中检索得到的1840种成分分子描述符进行主成分分析后,由得分图和载荷图结果可知分子描述符可以区分出溶解度有差异的化学成分类别;随机划分校正集和验证集后,基于化学成分群加和性分子描述符预测溶液浊度值的PLS预测模型的R²_Xcum=0.873,R²_Ycum=0.998,Q²cum=0.869;川芎、荆芥和益母草的浊度预测误差分别为3.4%、9.3%、23.8%。结论 基于化学成分群加和性分子描述符的中药浸膏粉溶化性预测模型具有一定的预测能力,研究结果为从中药整体成分层面预测宏观物性提供参考。

Objective The prediction model of the solubility of traditional Chinese medicine extract powder was developed and verified from the additive molecular descriptors of chemical ingredients group. Methods The in-situ turbidity sensor was used to evaluate the solubility of traditional Chinese medicine extract powder, and the test method conditions were optimized and verified. Taking extract powders of 13 representative traditional Chinese medicine [Guanghuoxiang (Pogostemonis Herba), Danggui (Angelicae Sinensis Radix), Bohe (Menthae Haplocalycis Herba), Qumai (Dianthi Herba), Yinchaihu (Stellariae Radix), Wumei (Mume Fructus), Gegen (Puerariae Lobatae Radix), Xixin (Asari Radix et Rhizoma), Yimucao (Leonuri Herba), Kushen (Sophorae Flavescentis Radix), Shudihuang (Rehmanniae Radix Praeparata), Jingjie (Schizonepetae Herba), and Chuanxiong (Chuanxiong Rhizoma)] as the research objects, the chemical components contained in each traditional Chinese medicine were retrieved, and the average properties of the chemical components of each traditional Chinese medicine were calculated according to the molecular descriptors of the components. Then, the partial least squares (PLS) method was used to establish and verify the correlation prediction model between the average property and the turbidity value of the corresponding traditional Chinese medicine solution. Results The objective evaluation method of the solubility of traditional Chinese medicine extract powder had good repeatability and durability. After principal component analysis of 1840 components molecular descriptors retrieved from 13 traditional Chinese medicines, it can be seen from the score plot and the load plot that the molecular descriptors can distinguish the chemical component categories with different solubility. The calibration set and the validation set were randomly divided, and the PLS prediction model of the solution turbidity value predicted based on the additive molecular descriptor of the chemical composition group had R²_Xcum=0.873,R²_Ycum=0.998,Q²cum=0.869. The turbidity prediction errors of Chuanxiong Rhizoma, Schizonepetae Herba and Leonuri Herba were 3.4%, 9.3% and 23.8%, respectively. Conclusion The solubility prediction model of traditional Chinese medicine extract powder from the additive molecular descriptors of chemical ingredients group had certain predictive ability, and the research results provided a reference for predicting macroscopic physical properties at the holistic composition level of traditional Chinese medicine.

R283.6

国家自然科学基金项目（82074033）；创新药物与高效节能降耗制药设备国家重点实验室开放基金项目（GZSYS202007）；连云港市重大技术攻关“揭榜挂帅”项目（CGJBGS2101）