Fu Xiao-Hua, Ye Yi-Fang, Luo Mei-Juan, Hong Xiao-Dan, Chen Xiao-Lu, Yao Qiu-Yan, Rong Ying-Ci, Ren Bin
Guangzhou Xinhai Hospital, Guangzhou 510300, China.
Yao Xue Xue Bao. 2012 Sep;47(9):1134-40.
This study is to establish an artificial neural network (ANN) for predicting blood tacrolimus concentration in liver transplantation recipients. Tacrolimus concentration samples (176 samples) from 37 Chinese liver transplantation recipients were collected. ANN established after network parameters were optimized by using momentum method combined with genetic algorithm. Furthermore, the performance of ANN was compared with that of multiple linear regression (MLR). When using accumulated dose of 4 days before therapeutic drug monitoring (TDM) of tacrolimus concentration as input factor, mean prediction error and mean absolute prediction error of ANN were 0.02 +/- 2.40 ng x mL(-1) and 1.93 +/- 1.37 ng x mL(-1), respectively. The absolute prediction error of 84.6% of testing data sets was less than 3.0 ng x mL(-1). Accuracy and precision of ANN are superior to those of MLR. The correlation, accuracy and precision of ANN are good enough to predict blood tacrolimus concentration.
本研究旨在建立一个用于预测肝移植受者血液中他克莫司浓度的人工神经网络（ANN）。收集了37名中国肝移植受者的他克莫司浓度样本（176个样本）。通过使用动量法结合遗传算法对网络参数进行优化后建立了人工神经网络。此外，将人工神经网络的性能与多元线性回归（MLR）的性能进行了比较。当将他克莫司浓度治疗药物监测（TDM）前4天的累积剂量作为输入因子时，人工神经网络的平均预测误差和平均绝对预测误差分别为0.02±2.40 ng x mL(-1)和1.93±1.37 ng x mL(-1)。84.6%的测试数据集的绝对预测误差小于3.0 ng x mL(-1)。人工神经网络的准确性和精密度优于多元线性回归。人工神经网络的相关性、准确性和精密度足以预测血液中他克莫司的浓度。
