设为首页 联系我们 加入收藏

当前位置: 网站首页 期刊分类目录 2025年第4期 正文

基于多头注意力机制和双向长短时记忆网络的DNA结合蛋白和RNA结合蛋白预测

作者:时间:2025-08-06点击数:



全文下载: 202504005.pdf


文章编号: 1672-6987(2025)04-0037-09DOI: 10.16351/j.1672-6987.2025.04.005



韦芹芹a 于彬b,c 张岩a*(青岛科技大学 a.数理学院; b.数据科学学院; c.信息科学技术学院, 山东 青岛 266061)

摘要: 提出了一种预测DNA结合蛋白(DBPs)和RNA结合蛋白(RBPs)的方法BiLSTM-MHA。首先,采用组成、转化和分布(composition, transition and distribution, CTD)、二肽偏离预期平均值(dipeptide deviation from expected mean, DDE)、二肽组成(dipeptide composition, DPC)、分组三肽组成(grouping tripeptide composition, GTPC)、伪氨基酸组成 (pseudo amino acid composition, PseAAC) 5种方法提取蛋白质序列的信息。进而将5种方法提取的特征向量进行融合。其次,利用组套索(group lasso)方法降低融合特征的维数,去除无关特征,提高模型预测准确率。最后,将多头注意力机制和双向长短时记忆网络结合,用于预测DNA和RNA结合蛋白预测。在十折交叉验证下,与其他已发表的模型进行比较,并在测试集上与其他方法进行对比。训练集和测试集的预测结果表明,所提出的BiLSTM-MHA模型能有效预测DBPs和RBPs。


关键词: DNA和RNA结合蛋白; 组套索; 多头注意力机制; 双向长短时记忆网络


中图分类号: Q 811.4        文献标志码: A


引用格式: 韦芹芹, 于彬, 张岩. 基于多头注意力机制和双向长短时记忆网络的DNA结合蛋白和RNA结合蛋白预测[J]. 青岛科技大学学报(自然科学版), 2025, 46(4): 37-45.


WEI Qinqin, YU Bin, ZHANG Yan. Predicting DNA- and RNA-binding proteins based on multi-head attention and BiLSTM[J]. Journal of Qingdao University of Science and Technology(Natural Science Edition), 2025, 46(4): -.

Predicting DNA- and RNA-Binding Proteins Based on Multi-Head Attention and BiLSTM

WEI Qinqina YU Binb,c ZHANG Yanaa.College of Mathematics and Physics; b.College of Data Science; c.College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, China)

Abstract: In this work, we construct a model to predict DNA binding proteins (DBPs) and RNA binding proteins (RBPs), called BiLSTM-MHA. First, five methods are used to extract protein sequence information, namely composition, transformation and distribution (CTD), dipeptide deviation from expected mean (DDE), dipeptide composition (DPC), grouping tripeptide composition (GTPC), and pseudo amino acid composition (PseAAC). Then the feature vectors extracted by the five methods are fused. Secondly, group lasso is used to reduce the dimension of fused features, remove irrelevant features, and improve the accuracy of model prediction. Finally, the multi-head attention mechanism is combined with Bi-directional Long Short-Term Memory to predict DBPs and RBPs. Under 10-fold cross-validation, BiLSTM-MHA is compared with other published models on the training datset, and compared with other methods on the test set. The prediction results show that the BiLSTM-MHA model proposed in this paper can effectively predict DBPs and RBPs.


Key words: DBPs and RBPs; group lasso; multi-head attention mechanism; Bi-directional long short-term memory

收稿日期: 2024-08-16

基金项目: 国家自然科学基金项目(62172248); 山东省自然科学基金项目(ZR2021MF098).

作者简介: 韦芹芹(1998—), 女, 硕士研究生.     * 通信联系人


Copyright © 2011-2017 青岛科技大学学报 (自然科学版)