人工神经网络预测HPMC缓释片中易溶性药物的释放

目的应用BP人工神经网络模型预测水溶性药物从HPMC缓释片中的释放。方法以6种不同溶解性的水溶性药物(对乙酰氨基酚、氧氟沙星、盐酸环丙沙星、乳酸左氧氟沙星、多索茶碱、氯苯那敏、维拉帕米)为模型药物,设计62个处方,其中前面55个处方作为训练处方,另外7个处方作为验证处方,压制HPMC缓释片,进行释放度检查。以溶解度、载药量、HPMC的量、HPMC的固有黏度、MCC的量、PVP的浓度和药物溶出仪的转速作为自变量,药物在各个取样时间点的累积释放量作为输出,建立BP人工神经网络模型,并与响应面法进行对照,通过线性回归法和相似因子法比较人工神经网络和响应面法的预测能力,借助三维图说明各个变量对药物释放的影响。结果线性回归和相似因子法表明人工神经网络较响应面法的预测值与实际测定值更吻合,更能充分地说明单因素对药物释放的影响规律。结论人工神经网络可以代替响应面法处理HPMC缓释片处方设计中的不同溶解度的水溶性药物的多因素多响应的非线性问题而且可以推广到别的制剂设计中。

An Artificial Neural Network to Predict Water Soluble Drug Release from HPMC Sustained Release Tablet

OBJECTIVE To use an artificial neural network(ANN) to predict water soluble drug release from HPMC matrix tablets.METHODS 62 formulations of six water soluble model drugs with different solubility(atlachlor,verapamil,hydrochloric ciprofloxacin,doxofylline, paracetamol,lactic acid Levofloxa) were prepared.The solubility of drug,the amount of loading drug,the amount of HPMC,the intrinsic viscosity of HPMC,the amount of MCC and the stirring speed of dissolution machine were regarded as casual variable(inputs) while the accumulated drug release in each sampling time were used as response variables(outputs),The 55 formulation in 62 preparation were used as trained data set,while the other formulations were used as the validating data sets(response variables),The series of response variables and casual variables were used as tutorial data for ANN to train BP network and use the trained network to predict drug release.Combined with 3-dimension diagram,we illustrated the influence of each variable on drug release. RESULTS The linenr regression equation parameter and the seminary factors showed,compared with RSM,the predicted values by ANNs were closer to the observed values both of the training data set and the validating data set.CONCLUSION ANN is an efficient way to deal with multi-variable-multi-objective question in designing HPMC sustained release tablets formulation with diffirent water-soluble drug,and it can be widely applied in formulation design.