基于碱基关联二联体位置权重矩阵预测酵母转录因子结合位点

Recognition of the Transcription Factor Binding Sites in Saccharomyces cerevisiae Genome Based on Dinucleotides Position Weight Matrix

基于已知的酵母转录因子结合位点数据资料,构建转录因子结合位点碱基关联二联体位置权重矩阵,整合碱基关联二联体位置权重矩阵和碱基保守性参量M2i,提出一种新的预测转录因子结合位点的方法(PWMSA).利用self-consistency和cross-validation两种方法对此算法进行检验,均获得了较高的预测成功率,结果表明9种转录因子结合位点的总体预测成功率超过81%,明显高于单碱基位置权重矩阵,同时与已有预测转录因子结合位点的软件进行比较,核苷酸水平上的关联系数和结合位点水平上的关联系数分别达到0.42和0.52,优于现有预测方法.

Based on the known transcription factor binding sites in Saccharomyces cerevisiae genome, a dinucleotides position weight matrix for transcription factor binding sites is constructed.By calculating the site conservative index vectors Mli in transcription factor binding sites, a novel position weight matrices scoring algorithm （PWMSA） for predicting yeast transcription factor binding sites is presented. The 9 yeast transcription factor binding sites sets which were confirmed by experiment are used to train this algorithm. The predictive capacity of the algorithm is tested by the 10-fold cross-validation test. The results show that the correct prediction is 81.1% more than mononucleotide PWM. By comparing our algorithm with other ten softwares using the new performance measures and benchmarked database, the results show that the overall prediction accuracies of PWMSA are 0.52 and 0.42 more than the other ten algorithms, at binding sites segment level and nucleotide level, respectively.