论文部分内容阅读
建立人工神经网络用于估算霉酚酸(MPA)体内暴露药量(AUC)。64例肾移植受者术后不同时间服用霉酚酸酯(MMF),于服药前以及服药后0.5、1、1.5、2、3、4、6、8和12h等10个时间点采取外周静脉血,采用高效液相色谱法检测血浆MPA浓度,用线性梯形法计算服药后0～12h药-时曲线下面积(AUC0-12h),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。以0、0.5、2h血药浓度数据预测AUC0-12h,人工神经网络平均预测误差(MPE)与平均绝对误差(MAE)分别为-1.53%和9.12%,准确度及精密度优于多元线性回归。以0、0.5h血药浓度数据预测AUC0-12h,人工神经网络MPE与MAE分别为6.03%和15.30%,准确度及精密度亦优于多元线性回归。人工神经网络预测的准确度和精密度均优于多元线性回归法,可用于预测MPAAUC0-12h。
An artificial neural network was established to estimate mycophenolic acid (MPA) in vivo exposure dose (AUC). Sixty-four renal transplant recipients received myofolate (MMF) at different times after operation. Peripheral venous blood was collected before and at 0.5, 1, 1.5, 2, 3, 4, 6, 8 and 12 h Plasma was collected to measure plasma MPA concentration by high performance liquid chromatography. The area under the drug-time curve (AUC0-12h) was calculated by linear trapezoidal method from 0 to 12 hours after drug administration. The artificial neural network was established by using genetic algorithm and momentum method to optimize the network parameters. Predicting AUC0-12h with 0, 0.5, 2h blood concentration data, the mean prediction error (MPE) and mean absolute error (MAE) of artificial neural network were -1.53% and 9.12% respectively, accuracy and precision were better than multivariate linear regression . Predicting AUC0-12h with 0,0.5h plasma concentration data, the artificial neural networks MPE and MAE were 6.03% and 15.30%, respectively. The accuracy and precision were also better than multivariate linear regression. Artificial neural network prediction accuracy and precision are better than multiple linear regression method can be used to predict MPAAUC0-12h.