目的 采用近红外光谱(near infrared spectroscopy，NIRS)、中红外光谱(mid-infrared spectroscopy，MIRS)技术实现对热毒宁注射液制备过程中金银花浓缩过程绿原酸、新绿原酸、隐绿原酸、异绿原酸A、异绿原酸B、异绿原酸C、断氧化马钱子苷和固含量8个质控指标的含量预测，并对比2种技术的预测效果。方法 收集热毒宁注射液制备过程中金银花浓缩过程样本，进行NIRS、MIRS采集和含量测定，优选最佳光谱预处理方法和特征波段，采用偏最小二乘(partial least square，PLS)法建立8个质控指标的含量预测模型，并比较8个质控指标的NIRS、MIRS模型性能，得到8个最优含量预测模型，并对其进行外部验证。结果 NIRS技术对绿原酸、隐绿原酸、异绿原酸C、断氧化马钱子苷、固含量的预测效果更好，平均相对预测误差(average relative prediction error，ARPE)分别为1.57%、1.88%、4.13%、3.79%、0.94%，故选用NIRS模型作为这5个质控指标的最佳模型；MIRS技术对新绿原酸、异绿原酸A、异绿原酸B的预测效果更好，ARPE分别为4.00%、4.01%、2.32%，故选用MIRS模型作为这3个质控指标的最佳模型。经外部验证，绿原酸、新绿原酸、隐绿原酸、异绿原酸A、异绿原酸B、异绿原酸C、断氧化马钱子苷和固含量预测模型的ARPE分别为6.22%、7.94%、5.44%、5.05%、5.37%、7.44%、6.00%和2.74%。结论 建立了NIRS、MIRS技术对热毒宁注射液制备过程中金银花浓缩过程的含量预测方法，该方法快速简便、结果可靠，且MIRS技术较NIRS技术有更低的检测限。

Objective Using near infrared spectroscopy (NIRS) and mid-infrared spectroscopy (MIRS) technology to predict the content of eight quality control indicators during the concentration process of Jinyinhua (Lonicerae Japonicae Flos, LJF) in preparation of Reduning Injection (热毒宁注射液), namely chlorogenic acid, neochlorogenic acid, cryptochlorogenic acid, isochlorogenic acid A, isochlorogenic acid B, isochlorogenic acid C, secoxyloganin, solid content, and to compare the prediction results of the two techniques.d. Methods The samples collected during the concentrated process of Reduning Injection LJF were collected for NIRS and MIRS acquisition and content determination. The optimal spectral preprocessing method and characteristic band were selected, and the partial least squares method was used to establish the content prediction model of eight quality control indicators. And compare NIRS and MIRS model performance of eight quality control indicators. Eight optimal content prediction models were obtained and externally verified. Results The NIRS technique had better prediction effect on chlorogenic acid, cryptochlorogenic acid, isochlorogenic acid C, secoxyloganin and solid content, the average relative prediction error (ARPE) was 1.57%, 1.88%, 4.13%, 3.79% and 0.94%, respectively. The prediction effect of MIRS technology was better for neochlorogenic acid, isochlorogenic acid A and isochlorogenic acid B, the ARPE was 4.00%, 4.01% and 2.32%, respectively. Therefore, chlorogenic acid, cryptochlorogenic acid, isochlorogenic acid C, secoxyloganin and solid were selected by NIRS model; neochlorogenic acid, isochlorogenic acid A and isochlorogenic acid B were selected the MIRS model as the best model. Through external verification, the ARPEs of chlorogenic acid, neochlorogenic acid, cryptochlorogenic acid, isochlorogenic acid A, isochlorogenic acid B, isochlorogenic acid C, secoxyloganin and solid content prediction models were 6.22%, 7.94%, 5.44%, 5.05%, 5.37%, 7.44%, 6.00% and 2.74%, respectively. Conclusion The established method of NIRS and MIRS technology for the content prediction of LJF in Reduning Injection during the concentration process is fast, simple and reliable, and MIRS technology has lower detection limit than NIRS technology.

R283.6

2022年中央财政转移支付地方项目：基于重点研究室研究领域的中医药多学科研究能力提升项目—中药提取精制新技术