基于最优分割位点的蛋白质亚细胞位点预测方法

Prediction of protein subcellular location using optimal cleavage site

蛋白质的亚细胞位点信息有助于我们了解蛋白质的功能以及它们之间的相互作用,同时还可以为新药物的研发提供帮助。目前普遍采用的亚细胞位点预测方法主要是基于N端分选信号或氨基酸组分特征,但研究表明,单纯基于N端分选信号或氨基酸组分的方法都会丢失序列的序信息。为了克服此缺陷,本文提出了一种基于最优分割位点的蛋白质亚细胞位点预测方法。首先,把每条蛋白质序列分割为N端、中间和C端三部分,然后在每个子序列和整条序列中分别提取氨基酸组分、双肽组分和物理化学性质,最后我们把这些特征融合起来作为整条序列的特征。通过夹克刀检验,该方法在NNPSL数据集上得到的总体精度分别是87.8%和92.1%。

Protein subcellular locations has immediate relevance for understanding protein function and designing new drug.Present methods are mainly based on sorting signals or amino acid compositions.However,methods based solely on sorting signals or amino acid compositions may lose the sequence order information.To overcome the shortcomings,we divided each chain into three parts：N-terminal,middle,and C-terminal.Then,features were extracted from each part and the whole chain independently.These features are amino acid compositions,dipeptides,and stereochemical properties.Finally,features of different parts are combined and the combined features are used as features of the whole chain.By Jackknife test on the NNPSL dataset,our overall accuracies for prokaryotic and eukaryotic proteins are 87.8% and 92.1%,respectively.