摘要
文章提出了一种应用于T细胞表位预测的模块化神经网络,这种神经网络先用一个过滤模块将不可结合的蛋白质序列过滤掉,然后将可结合的蛋白质输入到各个分类模块中进行训练学习,最后根据三个分类模块的结果进行最大值判定输出最后结果。实验结果证明,用这种模块化神经网络结构对T细胞表位进行预测比单个BP神经网络具有更高的准确率和数据的自我组织及学习能力。
A modular neural network method is proposed in this paper used for prediction of T cell epitope.at first the binding protein sequence is left by passing filtering net module,which is trained to classify whether the samples being binding or not,then the classification net modules are used to sort different binding type,by training for binding samples.At last,the last result is max value of three classification net mudules's results.Comparing with the results of single BP neural network,the results by the proposed approach are more accurate and are more better to integrate new data and self-improve.used for predictions of T cell epitope.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第24期33-35,共3页
Computer Engineering and Applications
基金
国家自然科学基金项目(编号:C030205)资助