遗传算法误差反向传播人工神经网络预测阿立哌唑血药浓度

Prediction of aripiprazole blood concentration by GA-BP artificial neural network

目的构建基于遗传算法误差反向传播(GA-BP)人工神经网络的阿立哌唑(APZ)及其代谢产物脱氢阿立哌唑(DAPZ)血药浓度预测模型,为需要调整APZ使用剂量或不能进行APZ血药浓度监测的患者提供浓度预测模型。方法回顾性收集在2021年7月—2022年8月新疆维吾尔自治区人民医院就诊且规律服用APZ的174例患者的血药浓度资料,提取相关变量,采用Matlab R2018a编程软件,结合深度学习网络构建GA-BP人工神经网络预测模型,预测APZ+DAPZ血药浓度。结果GA-BP人工神经网络预测模型验证结果显示,35例验证组样本的预测结果与实测结果相比,平均预测误差为-0.0926,平均绝对误差为0.6895,35个预测误差均小于15%,小于15%的概率为100%,血药浓度的预测值与实测值之间的相关系数为0.997,预测结果较理想。结论GA-BP人工神经网络预测模型预测APZ+DAPZ血药浓度,可用于APZ的个体化给药。

Objective To construct a genetic algorithm back propagation(GA-BP)artificial neural network model for predicting the blood concentration of aripiprazole(APZ)and its metabolite dehydro-aripiprazole(DAPZ),and to provide a concentration prediction model for patients who need to adjust the dose of APZ or cannot monitor APZ blood concentration.Methods Blood drug concentration data were collected retrespectively from 174 patients who regularly took APZ in Xinjiang Uygur Autonomous Region People's Hospital from July 2021 to August 2022.Relevant variables were extracted,and GA-BP artificial neural network prediction model was constructed by Matlab R2018a programming software combined with deep learning network to predict blood drug concentration of APZ+DAPZ.Results The GA-BP artificial neural network prediction model showed that compared with the measured results,the average prediction error and the average absolute error of the 35 samples in the verification group were-0.0926 and 0.6895,respectively.The 35 prediction errors were all less than 15%,and the probability of less than 15%was 100%.The correlation coefficient between the predicted value and the measured value was 0.997,and the predicted result was ideal.Conclusion GA-BP artificial neural network prediction model can be used to predict the blood concentration of APZ+DAPZ and for individual drug administration of APZ.