English

应用科学学报 ›› 2005, Vol. 23 ›› Issue (2): 187-191.

• 论文 • 上一篇    下一篇

ANN-CE:一种预测DNA结合位点的改进神经网络方法

徐东, 王翼飞   

  1. 上海大学理学院, 上海 200444
  • 收稿日期:2003-12-15 修回日期:2004-02-16 出版日期:2005-03-31 发布日期:2005-03-31
  • 作者简介:徐东(1978-),男,上海人,博士生,E-mail:xstone@citiz.net;王翼飞(1948-),男,上海人,教授,博导.
  • 基金资助:
    国家高技术研究发展计划(863计划)资助项目(2002AA234021)

ANN-CE: An Improved Neural Network Method for Predicting DNA Binding Sites

XU Dong, WANG Yi-fei   

  1. College of Sciences, Shanghai University, Shanghai 200444, China
  • Received:2003-12-15 Revised:2004-02-16 Online:2005-03-31 Published:2005-03-31

PDF

95

被引次数

1

摘要： 基于误差平方和最小原则的神经网络方法并不适于解决DNA结合位点的预测问题,提出了一种改进的神经网络方法(ANN-CE)被用于对DNA结合位点进行预测.这是一个以交叉熵函数为目标函数的三层反向传播神经网络.计算结果表明,与基于误差平方和最小原则的同规模BP网络相比,其对DNA结合位点预测的敏感性Sn(sensitivity)和特异性Sp(specificity)可分别提高11.40%和11.91%.

关键词: 基因调控, DNA结合位点, BP神经网络, 交叉熵

Abstract: The neural network technology has achieved great success in the field of bio-sequence analysis.But a neural network based on MSSE is not the final solution to the problem of predicting DNA binding sites.In this paper, An improved neural network method, ANN-CE, to predict DNA binding sites is presented.It is a three layered back-propagation neural network in which the error function is a cross entropy function.The result shows that the predicted sensitivity and specificity of ANN-CE increased by 11.40% and 11.91% respectively compared with the back-propagation neural network based on MSSE with the same architecture.

Key words: BP neural network, gene regulation, DNA binding site, cross entropy

中图分类号: