基于贝叶斯网的蛋白质相互作用位点预测

Interaction-site prediction for protein Based on Bayes Method

蛋白质相互作用位点研究在蛋白质功能分析及药物设计等方面有着重要的应用。文章以蛋白质中的氨基酸残基为研究对象,使用残基的溶剂可及表面积、进化保守性打分及残基的序列信息熵三个特征为特征集,构建了基于贝叶斯方法的蛋白质相互作用位点预测的贝叶斯分类预测器。方法有效的结合了蛋白质残基特征数据集经常性数据缺失的特点及贝叶斯网在处理不确定性数据方面的优点,通过对基准的71个蛋白质数据集进行实验,结果表明我们的分类器预测的有效性。

The research on protein-protein interaction siles has important application in the study of protein function analysis and drug design.Here we take the residue of protein as our research object,and take the accessible surface area of residue,residue conservation score and residue sequence entropy as our feature sets.A Bayesian classifier is constructed to infer protein interaction sites based on these feature sets.This method efficiently combined the characteristics of the protein's residue which often misses data and the characteristics of Bayesian network which has the advantage of processing uncertainty data.According this,we made tests on a benchmark dataset of 71 proteins.Experimental results showed the effectiveness of our method.