Wang Yan-Bin, You Zhu-Hong, Li Li-Ping, Huang Yu-An, Yi Hai-Cheng
Xinjiang Technical Institutes of Physics and Chemistry, Chinese Academy of Science, Urumqi 830011, China.
University of Chinese Academy of Sciences, Beijing 100049, China.
Molecules. 2017 Aug 18;22(8):1366. doi: 10.3390/molecules22081366.
Protein-protein interactions (PPIs) play a very large part in most cellular processes. Although a great deal of research has been devoted to detecting PPIs through high-throughput technologies, these methods are clearly expensive and cumbersome. Compared with the traditional experimental methods, computational methods have attracted much attention because of their good performance in detecting PPIs. In our work, a novel computational method named as PCVM-LM is proposed which combines the probabilistic classification vector machine (PCVM) model and Legendre moments (LMs) to predict PPIs from amino acid sequences. The improvement mainly comes from using the LMs to extract discriminatory information embedded in the position-specific scoring matrix (PSSM) combined with the PCVM classifier to implement prediction. The proposed method was evaluated on and datasets with five-fold cross-validation experiments. The experimental results show that the proposed method achieves high average accuracies of 96.37% and 93.48%, respectively, which are much better than other well-known methods. To further evaluate the proposed method, we also compared the proposed method with the state-of-the-art support vector machine (SVM) classifier and other existing methods on the same datasets. The comparison results clearly show that our method is better than the SVM-based method and other existing methods. The promising experimental results show the reliability and effectiveness of the proposed method, which can be a useful decision support tool for protein research.
蛋白质-蛋白质相互作用（PPIs）在大多数细胞过程中起着非常重要的作用。尽管已经投入了大量研究通过高通量技术来检测PPIs，但这些方法显然既昂贵又繁琐。与传统实验方法相比，计算方法因其在检测PPIs方面的良好性能而备受关注。在我们的工作中，提出了一种名为PCVM-LM的新型计算方法，该方法结合概率分类向量机（PCVM）模型和勒让德矩（LMs）从氨基酸序列预测PPIs。改进主要来自于使用勒让德矩提取嵌入在位置特异性得分矩阵（PSSM）中的判别信息，并结合PCVM分类器进行预测。所提出的方法在数据集上通过五折交叉验证实验进行了评估。实验结果表明，所提出的方法分别实现了96.37%和93.48%的高平均准确率，远优于其他知名方法。为了进一步评估所提出的方法，我们还在相同数据集上，将所提出的方法与最先进的支持向量机（SVM）分类器及其他现有方法进行了比较。比较结果清楚地表明，我们的方法优于基于SVM的方法和其他现有方法。这些有前景的实验结果表明了所提出方法的可靠性和有效性，它可以成为蛋白质研究中一个有用的决策支持工具。
