摘要
目的建立人工神经网络模型用于有限采样点估算霉酚酸(MPA)的药动学。方法 64例肾移植受者于术后不同时间服用霉酚酸酯(MMF)前以及服药后0.5、1、1.5、2、3、4、6、8、10、12 h等11个时点采取外周静脉血,采用高效液相色谱法测定血浆MPA浓度,采用遗传算法配合动量法优化网络参数,建立人工神经网络模型。结果以服药前及服用MMF 0.5、2 h后血药浓度数据预测MPA药动学,人工神经网络平均预测误差(MPE)与平均绝对误差(MAE)分别为(0.39±1.24)和(0.90±0.94)μg·mL-1。以MPA预测浓度计算的药动学参数与实际浓度计算的药动学参数无显著差异(P>0.05)。结论人工神经网络可用于有限采样点法预测MPA药动学过程。
ObjectiveTo establish an artificial neural network （ANN） for predicting mycophenolic acid （MPA） pharmacokinetics in renal transplantation recipients by limited sampling strategy. Methods Sixty-four Chinese renal transplantation recipients receiving mycophenolate mofetil （MMF） were investigated. Eleven serum samples were drawn on different days after transplantation. MPA plasma concentration was determined by HPLC and ANN was established after the network parameters were optimized by using momentum method combined with genetic algorithm. Results When using MPA plasma concentrations at 0, 0.5 and 2 h after MMF administration to predict MPA concentration, the mean prediction error and mean absolute prediction error were （0.39±1.24） and （0.90±0.94） μg·mL-1, respectively. There was no significant difference between the pharmacokinetic parameters obtained from the predicted MPA concentrations vs those obtained from the measured MPA concentrations. CONCLUSON ANN can be used to predict MPA pharmacokinetics by limited sampling strategy.
出处
《中国药学杂志》
CAS
CSCD
北大核心
2013年第14期1200-1203,共4页
Chinese Pharmaceutical Journal
基金
广东省医院药学研究基金课题(200731)
关键词
霉酚酸
药动学
肾移植
人工神经网络
mycophenolic acid, pharmacokinetics, renal transplantation, artificial neural netword
