毛细管电泳径向基神经网络校正法定量分析核苷  

Calibration Modeling of Capillary Electrophoresis for Quantitative Analysis of Guanosine Based on Radial Basis Function Network

作　　者：毛利锋[1] 沈朋[2] 程翼宇[1] 

机构地区：[1]浙江大学药物信息学研究所,杭州310027 [2]浙江大学医学院附属第一医院,杭州310003

出　　处：《化学学报》2004年第19期1917-1921,共5页Acta Chimica Sinica

基　　金：国家 863计划 (No.2 0 0 3AA2Z2 0 0 2 );浙江省科技计划 (No.2 0 0 4C330 2 6)资助项目

摘　　要：采用径向基神经网络算法对一组已知样品的核苷及内标物浓度与毛细管电泳峰面积数据进行回归计算 ,建立峰面积与核苷浓度之间的关系模型 ,对未知样品中待测核苷浓度作出预测 ,形成了毛细管电泳定量分析新方法 .将其用于鸟嘌呤核苷含量测定 ,所建模型预测结果平均相对误差为 0 86% ,明显低于线性回归及BP神经网络模型的 2 60 %和 1 0 7% .研究结果表明 ,本方法简便易用 ,能有效提高毛细管电泳定量分析的准确度 ,优于线性回归及BP神经网络法 .A new method for calibration modeling of capillary electrophoresis (CE) is a nonlinear regression based on radial basis function network (RBFN). The RBFN was trained with a set of known concentration data and CE peak area of guanosine and internal standards to establish the calibration model between the concentration of guanosine and its peak area. Subsequently, the RBFN model established can predict the concentration of guanosine by the peak area of guanosine and internal standards. For comparison, the calibration modeling of CE was carried out using a linear regression equation, BP artificial neural network ( BP-ANN) and RBFN approaches. The results of predicting the concentration of guanosine showed that the RBFN model had a lower mean relative error of prediction (0.86%) than those of linear regression equation (2.60%) and BP-ANN model (1.07%). The method is easy to be used and can effectively improve the accuracy of CE quantitative analysis.

关 键 词：毛细管电泳 径向基神经网络 校正方法 定量分析 核苷 

分 类 号：O658.9[理学—分析化学]