利用序列保守模体和局部构象信息预测转录因子结合位点  

Computational Prediction of Transcription Factor Binding Sites Based on Conserved Motif and Local Conformational Knowledge in Genomic Sequences

作　　者：杜耀华[1] 倪青山[1] 王正志[1] 

机构地区：[1]国防科技大学机电工程与自动化学院,中国湖南长沙410073

出　　处：《生命科学研究》2006年第3期215-223,共9页Life Science Research

基　　金：国家自然科学基金资助项目(60471003)

摘　　要：转录因子结合位点的计算预测是研究基因转录调控的重要环节,但常用的位置特异得分矩阵方法预测特异性偏低.通过深入分析结合位点的生物特征,提出了一种综合利用序列保守模体和局部构象信息的结合位点预测方法,以极大相关得分矩阵作为保守模体的描述模型,并根据二苷参数模型计算位点序列的局部构象,将两类信息得分组合为多维特征向量,在二次判别分析的框架下进行训练和滑动预测.预测过程中还引入了位置信息量以优化似然得分和过滤备选结果.针对大肠杆菌CRP和Fis结合位点数据的留一法测试结果表明,描述模型的改进和多种信息的融合能有效地改善预测方法的性能,大幅度提高特异性.Computational prediction of transcription factor binding sites （TFBSs） is an essential task in the research of transcription regulation, but the current algorithms using position specific scoring matrices （PSSMs） are limited in specificity. According to the thorough analysis about biologic features of TFBSs, a new prediction algorithm integrating conserved motifs and local conformational knowledge is proposed. Maximal dependence scoring matrices （MDSMs） are designed to model the motifs, and the dinucleotide step parameters are adopted to calculate the local conformation. Then the two kinds of feature scores are combined as muhidimensional vectors. Based on such vectors, the quadratic discriminant classifier can be trained to predict putative TFBSs through a sliding window. During the prediction, the positional information content is also introduced to optimize the scoring function and pre-select the potential matches. The results of leave-one-out tests on E. eoli CRP and Fis binding sites show that the algorithm can predict the TFBSs efficiently. Comparing with the PSSM based algorithm, it greatly decreases the number of false positives and improves the specificity.

关 键 词：转录因子结合位点 计算预测 保守模体 极大相关得分矩阵 局部构象 二次判别分析 

分 类 号：Q527[生物学—生物化学]