Ren Bin, He Qiu-Yi, Xu Qiong, Wang Chang-Xi, Chen Jie, Zheng Zhi-Hua, Li Shu-Xia, Chen Xiao
The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou 510080, China.
Yao Xue Xue Bao. 2009 Dec;44(12):1397-401.
The paper is aimed to establish an artificial neural network (ANN) for predicting mycophenolic acid (MPA) area under the plasma concentration-time curve (AUC) in renal transplantation recipients. 64 Chinese renal transplantation recipients receiving mycophenolate mofetil (MMF) were investigated. 10 timed samples were drawn at different days after transplantation. Plasma MPA concentration was determined by HPLC method and area under curve over the period of 0 to 12 h (AUC(0-12 h)) was calculated using the linear trapezoidal rule. ANN was established after network parameters were optimized using momentum method in combination with genetic algorithm. Furthermore, the predictive performance of ANN was compared with that of multiple linear regression (MLR). When using plasma MPA concentration of 0, 0.5, 2 h after MMF administration to predict MPA AUC(0-12 h), mean prediction error and mean absolute prediction error were -1.53% and 9.12%, respectively. Accuracy and precision of prediction by ANN were superior to that of MLR prediction, and similar results could be found when using plasma MPA concentration of 0, 0.5 h to predict MPA AUC(0-12h). The accuracy and precision of ANN prediction were superior to that of MLR prediction, and ANN can be used to predict MPA AUC(0-12 h).
本文旨在建立一种人工神经网络（ANN），用于预测肾移植受者血浆中霉酚酸（MPA）浓度-时间曲线下面积（AUC）。对64例接受霉酚酸酯（MMF）治疗的中国肾移植受者进行了研究。在移植后的不同日期采集10次定时样本。采用高效液相色谱法测定血浆MPA浓度，并使用线性梯形法则计算0至12小时期间的曲线下面积（AUC(0-12 h)）。采用动量法结合遗传算法对网络参数进行优化后建立了ANN。此外，还将ANN的预测性能与多元线性回归（MLR）的预测性能进行了比较。当使用MMF给药后0、0.5、2小时的血浆MPA浓度预测MPA AUC(0-12 h)时，平均预测误差和平均绝对预测误差分别为-1.53%和9.12%。ANN预测的准确性和精密度优于MLR预测，当使用0、0.5小时的血浆MPA浓度预测MPA AUC(0-12h)时也能得到类似结果。ANN预测的准确性和精密度优于MLR预测，且ANN可用于预测MPA AUC(0-12 h)。
