Abstract: In order to improve the prediction accuracy of protein secondary structure, a cascade neural networks composed of two-level network is presented. The first level is composed of five subnets with different structure, and the coding method of the second-level is studied and improved. The model is employed to predict 36 nonhomologous protein sequences with 6 122 residues in PDBSelect25. Results show that the proposed model can efficiently improve the prediction accuracy, increasing the prediction accuracy by 5.31%, 1.21% and 0.92% respectively compared with SNN, DSC and PREDSATOR method, improving the average prediction accuracy to 69.61%.

Key words: neural networks, protein, secondary structure prediction

摘要： 为提高蛋白质二级结构预测的精度，提出一种由两层网络构成的级联神经网络模型。第1层网络采用具有差异度的5个子网构成的网络模型，对第2层网络的输入编码进行改进。对PDBSelect25中的36条蛋白质共6 122个残基进行测试，结果表明，该模型能有效预测蛋白质二级结构，其预测精度分别比SNN, DSC, PREDSATOR方法提高5.31%, 1.21%和0.92%，平均预测精度提高到69.61%。

关键词: 神经网络, 蛋白质, 二级结构预测

CLC Number: 

• TP183