Prediction of oral hepatotoxic dose of natural products derived from traditional Chinese medicines based on SVM classifier and PBPK modeling
Prediction of oral hepatotoxic dose of natural products derived from traditional Chinese medicines based on SVM classifier and PBPK modeling
复制标题
DOI：
10.1007/s00204-021-03023-1
复制
发表时间：
2021-03-13
影响因子：
6.1
通讯作者：
Xiang, Xiaoqiang
中科院分区：
文献类型：
--
作者：
Li, Size;Yu, Yiqun;Xiang, Xiaoqiang
药物性肝损伤（DILI）的风险是中药天然产物（np - tcm）开发的主要挑战。寻找一种新的评价np -中药安全性的方法迫在眉睫。最近的研究报道了一种体外/计算机方法，使用支持向量机（SVM）分类器和基于生理的药代动力学（PBPK）建模来估计可接受的肝毒性化合物的每日摄入量。然而，该方法不适用于估计以多个日剂量给药的化合物的给药时间表。因此，在本研究中，对上述方法进行了特别优化，并用于估计17种np -中药的肝毒性血药浓度。此外，通过支持向量机分类器和PBPK模型预测雷公藤甲素、大黄素、苦参碱和氧化苦参碱的口服给药方案。优化包括：(1)采用基准浓度（BMC）模型对28个训练集化合物的体外细胞毒性数据进行优化；(2)采用训练集化合物的AUC代替C-max作为体内指标，可以更好地反映多剂量给药化合物的日暴露总量；(3)以血浆平均AUC为体内指标，以BMC值为体外指标，可获得较好的毒性分离指标（0.962比0.938）；(4)本研究计算的C-max和BMC值的TSI为0.985，结果表明BMC建模提高了分离性能。该优化的体外外推（IVIVE）工作流程可以推断出体外BMC对血药浓度和口服给药计划的影响，这些影响与一定的肝毒性风险相对应。该方法估计的氧化苦参碱安全给药方案与临床推荐给药方案接近。结果表明，优化后的方法可用于预测多剂量给药的给药计划，优化后的工作流程可用于np -中药的安全性评价和研发。
The risk of drug-induced liver injury (DILI) poses a major challenge for development of natural products derived from traditional Chinese medicines (NP-TCMs). It is urgent to find a new method for the safety assessment of the NP-TCMs. Recent study has reported an in vitro/in silico method to estimate the acceptable daily intake of hepatotoxic compounds using support vector machine (SVM) classifier and physiologically based pharmacokinetic (PBPK) modeling. However, this method is not suitable for estimating the dosing schedule of compounds which are administered in multiple daily doses. Thus, in this study, the method mentioned above was in particular optimized, and used to estimate the hepatotoxic plasma concentrations of 17 NP-TCMs. Additionally, the oral dosing schedules of the triptolide, emodin, matrine and oxymatrine were also predicted by the SVM classifier and PBPK modeling. The optimization included that: (1) in vitro cytotoxicity data of 28 training set compounds was optimized using benchmark concentrations (BMC) modeling; (2) AUC of the training set compound was used as the in vivo metric instead of C-max to better reflect the total daily exposure of compounds which are administered in multiple daily doses; (3) using the mean AUC in plasma as in vivo metric and BMC value as in vitro metric could achieve the better toxicity separation index (0.962 vs. 0.938); (4) The TSI for C-max and BMC values was 0.985 calculated in this study, and the results indicated that BMC modeling improved the separation performance. This optimized in vitro-in vivo extrapolation (IVIVE) workflow could extrapolate in vitro BMC to blood concentrations and the oral dosing schedule which are corresponding to certain risk of hepatotoxicity. The estimated safe dosing schedule of oxymatrine by this optimized method was close to the clinical recommended dosing regimen. The results indicate that the optimized method could be used to predict the dosing schedule of compounds administered in multiple daily doses, and our optimized workflow could be helpful for the safety assessment as well as the research and development on NP-TCMs.