【作者】 宋亮； 张剑； 王先明； 温继敏； 陈海波； 姚端正； 蒋昌忠；

【Author】 SongLiang Zhang Jian Wang Xianming Wen Jimin Chen Haibo Yao Duanzheng Jiang Changzhong (School of Physics Science and Technology, Wuhan University, Wuhan 430072)

【机构】 武汉大学物理科学与技术学院；

【摘要】 本文主要论述了神经网络在预测蛋白质二级结构方面的进展以及存在的问题。给出了神经网络预测二级结构的一般方法,指出了在二级结构指定,信息提取及训练几率方面存在的问题。研究了二级结构指定、训练几率与预测精度之间的关系。针对问题,提出了同时改进二级结构指定和调整三种残基训练几率的解决办法。更多还原

【Abstract】 Progress and problems in the field of predicting protein secondary structure by neural networks are presented in this paper. Especially, the problems on protein secondary structure assignment, probability of training for the three categories and information extraction are discussed. We present here the relationship between those problems and the accuracy of the result as well as the common method of protein secondary structure prediction by neural networks. Combining the advantage of the improvement of protein secondary structure assignment and proper modification for the probability of training will generate a better result.更多还原