基于小波分析的BP神经网络膜蛋白跨膜螺旋区段预测

基因组计划所产生的大量蛋白质序列迫切需要从理论上预测跨膜螺旋区段。提出了基于小波多分辨分析的BP神经网络膜蛋白跨膜螺旋区段的预测新方法,并把此方法称之为WnnTM。从MPtopo数据库中随机抽取80条三维结构已知的膜蛋白质序列构建数据集,把它们映射成疏水值序列,通过小波分解和重构得到小波系数,并结合BP神经网络构造小波BP神经网络预测模型,对膜蛋白跨膜螺旋区段的位置和数目进行预测。实例验证,WnnTM预测方法比单独用BP神经网络对膜蛋白跨膜螺旋区段进行预测更有效。

Prediction of Transmembrane Helical Segments in Membrane Proteins Based on Back Propagation Neural Network of Wavelet Analysis

The increasing protein sequences from the genome project require theoretical methods to predict transmembrane helical segments(TMHs).In this paper,WnnTM,a new method of BP neural network based on wavelet multiresolution analysis(MRA) is initially developed to predict the number and location of TMHs in membrane proteins.Eighty proteins with known 3D-structure chosen randomly from Mptopo database are mapped onto sequences of hydrophobicity values.Combining BP neural network with wavelet coefficients,which are through decomposing and reconstructing wavelet,the WnnTM method was constructed to predict the location and number of TMHs in membrane proteins.The obtained results indicate that the presented method is more effective than BP neural networks.