基于最大熵模型预测蛋白质结构的分类

基于最大熵模型,构建一种简单的预测蛋白质序列结构分类的算法。不同性质的氨基酸组合,在特定结构的蛋白质二级结构中,出现的频率不同,通过在模体数据库Prosite中查找蛋白质序列匹配的模体,以10种氨基酸组合在序列中出现的频率,表示蛋白质序列的特征,构建相应的结构分类预测模型。最大熵模型用来确定蛋白质结构分类预测模型的参数。以自身一致性和Jackknife测试方法验证分类模型的准确性。结果表明新构建的方法简单、准确,综合性能优于一般的预测方法。

Prediction of protein structure class with maximum entropy model

A simple and novel maximum entropy-based method is introduced to predict protein structure class.The motifs in protein se- quence are achieved through scanning protein sequence in protein motif dataset(Prosite).According to the segments of amino with dif- ferent characteristics occurring in protein sequence,the frequencies of 10 kinds of segments of amino acid in protein are calculated. The results of prosite and the number of 10 kinds of motifs within sequence are combined to represent the features of protein sequence. Maximum entropy model approach is used to determine the parameters of the predictive model.The new approach is evaluated on two benchmark dataset.Compared with prior works,the test results illuminate that the proposed approach is effective and promising.