基于PCA和BP神经网络的O—糖基化位点的预测和模式分析

Pattern analysis and prediction of O-linked glycosylation sites in protein based on PCA and BP neural networks

糖基化是真核细胞中最常见的翻译后蛋白质修饰过程之一。传统的神经网络方法已被应用预测蛋白质糖基化位点,预测的准确性主要依赖于特征向量的维数(蛋白质序列的长度),并随着蛋白质序列长度的增加而提高,但网络的结构变得越来越复杂,增加了计算运行成本。为了解决这一问题,提出了主成分分析和BP神经网络相结合的新方法对O—糖基化位点进行预测和分析,用PCA提取主成分构造子空间以降低输入的蛋白质序列的维数,再用BP神经网络预测一个特定的蛋白质序列是否被糖基化。实验表明,提出的新方法能大大缩短计算时间,并能提高预测的准确性。

Glycosylation is one of the most common post-translation modification of protein in eukaryotic cells.Conventional neural network method can be applied to predict glycosylation site in protein.The prediction accuracy mainly depends on the dimension of feature vector（the length of encode protein sequence）.But with the increasing of window size,the structure of neural network becomes more complex definitely and it is time-consuming.A new method based on principal component analysis（PCA） and BP（Back Propagation）neural network for pattern analysis and prediction O-linked glycosylation site in different windows size was proposed to solve this question.Firstly,PCA is applied to extract feature and reduce dimension,then,the neural network is used to predict whether a particular site of serine or threonine is glycosylated.Compared with conventional neural network methods,the proposed method shows that the performance has a better improvement,and the method can greatly save training time.