基于贝叶斯网络和相互作用可信度的蛋白质功能预测方法

Predicting Protein Function Based on Bayesian Network and Protein Interaction Reliability

蛋白质功能注释是后基因组时代研究的核心内容之一，基于蛋白质相互作用网络的蛋白质功能预测方法越来越受到研究者们的关注。提出了一种基于贝叶斯网络和蛋白质相互作用可信度的蛋白质功能预测方法。该方法在功能预测过程中为待注释的蛋白质建立贝叶斯网络预测模型，并充分考虑了蛋白质相互作用的可信度问题。在构建的芽殖酵母数据集上的三重交叉验证测试表明，在功能预测过程中考虑蛋白质可信度能够有效地提高功能预测的性能。与现有一些算法相比，该方法能够给出令人满意的预测效果。

Function annotation for proteins is one of the most important problems in the post-genomic era, and the protein-protein interaction data are employed by many researchers to assign functions to proteins. In this paper, a new method is developed to predict protein functions, which is based on Bayesian network and protein interaction reliability. The proposed algorithm constructs a Bayesian network model to assign functions to the unannotated protein and takes the reliability of protein interactions into account. The results, which is obtained by 3-fold cross-validation test on the data set constructed for Saccharomyces, show that the performance of protein function prediction can be improved by using interaction reliability, moreover, the proposed method outperforms some existing approaches and can be used to obtain desirable results for protein function prediction.