基于序列拓扑和二阶隐马尔可夫模型的跨膜蛋白亚细胞定位预测

现有蛋白质亚细胞定位方法针对水溶性蛋白质而设计,对跨膜蛋白并不适用。而专门的跨膜拓扑预测器,又不是为亚细胞定位而设计的。文章改进了跨膜拓扑预测器TMPHMMLoc的模型结构,设计了一个新的二阶隐马尔可夫模型;采用推广到二阶模型的Baum-Welch算法估计模型参数,并把将各个亚细胞位置建立的模型整合为一个预测器。数据集上测试结果表明,此方法性能显著优于针对可溶性蛋白设计的支持向量机方法和模糊k最邻近方法,也优于TMPHMMLoc中提出的隐马尔可夫模型方法,是一个有效的跨膜蛋白亚细胞定位预测方法。

Subcellular Localization Prediction of Transmembrane Proteins Based on Sequence Topology and Second-order Hidden Markov Model

Current predictors for subcellular localization primarily target soluble proteins and ignore the characteristic topological domains of transmembrane proteins.On the other hand,topology predictors are not designed for subcellular localization prediction.Inspired by the topology of transmembrane proteins,and based on the Hidden Markov Model(HMM) which was presented in a topology predictor named TMPHMMLoc,a modified second-order HMM was constructed for subcellular localization prediction of transmembrane proteins.An extended Baum-Welch algorithm was presented to estimate parameters of models. And a subcellular locations predictor was constructed by integrating all HMMs for every subcellular location in one model.529 transmembrane proteins locating at five subcellular sites of secretory pathways in mouse cells were extracted for screening and testing.In comparison with linear Support Vector Machines(SVMs) and Fuzzy k-Nearest Neighbors,based on overall amino acid and di-peptide composition,the method in this paper show a significant increase in prediction performance.And it also outperformed the HMM method of TMPHMMLoc.All these results indicate that the method in this paper is powerful for transmembrane proteins subcellular localization prediction.