蛋白质中不同类型β-转角的预测

Prediction of Different β-turn Types in Proteins

蛋白质二级结构预测是空间结构预测的关键步骤,β-转角是蛋白质二级结构中的重要类型之一,因此,找到一种准确预测不同类型β-转角的方法,对于建立蛋白质三维空间结构有重要意义。本文构建了一个含有2878条蛋白质链的数据库,比目前普遍使用的含有426条蛋白质链的数据库扩大了6倍多。以矩阵打分值、离散增量值、相似性值和预测的二级结构信息作为组合向量输入到支持向量机中,对不同β-转角进行了预测,各类的预测总精度均大于92.8%,MCC值均大于0.285。

Protein secondary structure prediction is a key step in overall tertiary structure prediction.β-turn is important component of protein secondary structure.Development of an accurate prediction method of β-turn types would be helpful for predicting the overall tertiary structure of proteins.We construct a database containing 2878 protein chains.It is almost 6 times bigger than the database of 426 protein chains that has been widely used in the prediction of β-turn types.We use the composite vector with matrix scoring value,increment of diversity value,similarity value and predicted secondary structure information as input parameter of the support vector machine algorithm to predict the β-turn types,and obtain the overall prediction accuracy exceed 92.8%,with the Matthews Correlation Coefficient values exceed 0.285 for types.