期刊
NUCLEIC ACIDS RESEARCH
卷 33, 期 -, 页码 W230-W232出版社
OXFORD UNIV PRESS
DOI: 10.1093/nar/gki412
关键词
-
Correctly predicting the disulfide bond topology in a protein is of crucial importance for the understanding of protein function and can be of great help for tertiary prediction methods. The web server http://clavius.bc. edu/similar to clotelab/ DiANNA/ outputs the disulfide connectivity prediction given input of a protein sequence. The following procedure is performed. First, PSIPRED is run to predict the protein's secondary structure, then PSIBLAST is run against the non-redundant SwissProt to obtain a multiple alignment of the input sequence. The predicted secondary structure and the profile arising from this alignment are used in the training phase of our neural network. Next, cysteine oxidation state is predicted, then each pair of cysteines in the protein sequence is assigned a likelihood of forming a disulfide bond-this is performed by means of a novel architecture (diresidue neural network). Finally, Rothberg's implementation of Gabow's maximum weighted matching algorithm is applied to diresidue neural network scores in order to produce the final connectivity prediction. Our novel neural network-based approach achieves results that are comparable and in some cases better than the current state-of-the-art methods.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据