使用双向LSTM的深度神经网络预测蛋白质残基相互作用

发布时间：2018-07-10 14:21

  本文选题：蛋白质 + 相互作用　； 参考：《小型微型计算机系统》2017年03期

【摘要】：残基对的相互作用描述了蛋白质三维结构中一对残基的空间距离关系.一对残基是否相互作用不仅取决于这对残基的本身属性,还受到这对残基所在蛋白质的所有其它残基的影响.传统的残基相互作用预测方法往往选取要预测残基对本身以及它们各自邻居的残基属性作为特征,这些方法忽略了影响残基对相互作用的全局因素.本文使用双向LSTM(Long Short-term M emory)抽取蛋白质序列上每个残基的属性,通过这种方式得到的每个残基属性不仅包含了局部属性还包含了全局属性.实验结果表明我们的模型在多个基准测试集上的Acc(Accuracy)超过其它方法 10%以上.
[Abstract]:The interaction of residue pairs describes the spatial distance relationship of a pair of residues in a protein three-dimensional structure. Whether a pair of residues interact or not depends not only on the properties of the pair of residues, but also on all the other residues of the protein in which the residues are located. The traditional prediction methods of residue interaction often choose the residual properties of the residual pair and their neighbors as the characteristics. These methods ignore the global factors that affect the residual pair interaction. In this paper, bidirectional LSTM (long Short-term M emory) is used to extract the attributes of each residue on a protein sequence. Each residue attribute obtained in this way contains not only local attributes but also global attributes. The experimental results show that our model has more than 10% Acc (Accuracy) on multiple benchmark sets.
【作者单位】： 苏州大学计算机科学与技术学院;苏州大学江苏省计算机信息处理技术重点实验室;
【基金】：国家自然科学基金项目(61170125)资助
【分类号】：Q51;TP183
，

本文编号：2113648