杨树蛋白质二级结构的人工神经网络预测

以PDB公用生物信息数据库为基础，用人工神经网络建模技术对杨树蛋白质二级结构预测模型进行了研究，这对深入认识森林动态机理、提高森林资源信息化管理水平、改进森林保护制药设计等具有重要的应用价值，对森林生物学、森林生物信息学的研究具有重要的学术意义。从公用数据库下载杨树蛋白质样本27个，提取氨基酸2 947个。用长度为17的滑动窗口截取蛋白质一级结构的氨基酸序列片段，并用[-1，1]编码方式进行编码，以组织输入向量，以片段中心氨基酸对应的蛋白质二级结构（螺旋、折叠、无规则卷曲）为输出向量，构建了结构为17∶S∶3的BP人工神经网络模型。用训练、测试样本对模型进行训练、检验，得出理想的模型结构为17∶9∶3，其总体拟合准确度为71%，总体预测准确度为65%，H的预测准确度达81%，比以往同类研究具有较高的预测准确度。

Prediction of Poplar Protein Secondary Structure with Artificial Neural Networks

The paper studied a model for predicting the secondary structure of poplar protein with artificial neural network modeling technology, based on the public biological information database, PDB that has important application value for understanding forest dynamic mechanism, and raising the level of forest resources information management, and improving pharmaceutical design of forest protection. It has important academic significance for studying forest biology and bioinformatics. Twenty seven poplar protein samples were downloaded from public database, 2947 amino acids were extracted. An amino acid sequence fragment of protein primary structure was cut out with a sliding window that length was 17, and was coded by [-1,1] coding scheme. The BP neural network model with the structure, 17∶S∶3, was created, by taking the [-1,1] coding as input variable, and protein secondary structure (spiral, folding, and random curl three state) corresponding amino acids at the center of the fragment as output variable. Through training and optimum seeking, the idea model structure was 17∶9∶3, the overall fitting accuracy was 71%, the overall prediction accuracy was 65%, and the prediction accuracy of H was 81%. The results indicated that the proposed model had higher prediction accuracy than the similar studies in the past.