基于神经网络的多聚脯氨酸二型结构预测

使用“机器学习方法”对稀有的多聚脯氨酸二型(PPII)的二级结构进行预测,在预处理蛋白质序列的基础上,使用生物信息学中常用的 BP神经网络预测 PPII二级结构.通过对不同输入窗口长度与不同隐层节点数的神经网络进行训练和测试,得出在输入窗口长度为 13 个氨基酸残基和隐节点数为15时预测效果最好,此时的预测精度可达73.8%.

Prediction of Polyproline Type Ⅱ Secondary Structures by Artificial Neural Network

So far few works have been conducted to predict polyproline type II(PPII) secondary structure with machine learning approaches.On the base of preprocessing protein sequences, this paper predicts PPII secondary structure with BP type's neural network model that is most frequently used in bioinformatics. By training and testing neural networks with different window lengths and different hidden nodes, it could be concluded that, the best predicting result was obtained when the window length was 13 and the the hidden node was 15,the optimal predicting accuracy(Q) reached 73.8 percent.