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 2024第5期 正文

基于双向长短时记忆网络和自注意力机制的药物-药物相互作用预测

作者:时间:2024-11-19点击数:



全文下载: 20240501.pdf


文章编号: 1672-6987202405-0149-10DOI10.16351/j.1672-6987.2024.05.020


张明香a, 顾海明a*,  于彬b*(青岛科技大学 a.数理学院; b.数据科学学院, 山东 青岛 266061)


摘要: 提出了一种基于双层双向长短时记忆网络(bi-directional long short term memoryBiLSTM)和自注意力(self-attention)机制的药物-药物相互作用(drug-drug interactionsDDIs)预测方法SA-BiLSTM。首先,利用FP3指纹、MACCS指纹、Pubchem指纹和PaDEL分子描述符对药物特征信息进行提取。其次,使用套索回归(least absolute shrinkage and selection operatorLasso)方法消除对分类无关的特征,并利用重复编辑最近邻(repeated edited nearest neighborsRENN)方法对数据进行平衡处理,得到最优特征向量。最后,将最优特征向量输入结合自注意力机制和双向长短时记忆网络的分类器预测DDIs。基于五折交叉验证,同时与其它预测方法进行比较,本工作所提出的方法在两个数据集上获得较高的预测准确率。为了综合评价SA-BiLSTM的性能,对药物-药物相互作用网络进行验证。实验结果表明,SA-BiLSTM表现出优秀的预测能力,可以为DDIs的预测提供一种新的思路。


关键词: 药物-药物相互作用; 特征提取; 重复编辑最近邻; 双向长短时记忆网络; 自注意力机制


中图分类号: Q 811.4文献标志码: A


引用格式: 张明香, 顾海明,  于彬. 基于双向长短时记忆网络和自注意力机制的药物-药物相互作用预测[J. 青岛科技大学学报(自然科学版), 2024, 45(5): 149-158.


ZHANG Mingxiang, GU Haiming, YU Bin. Predicting drug-drug interactions with BiLSTM and self-attention mechanismJ. Journal of Qingdao University of Science and TechnologyNatural Science Edition), 2024455): 149-158.


Predicting Drug-Drug Interactions with BiLSTM and Self-Attention Mechanism


ZHANG Mingxianga, GU Haiminga, YU Binb

(a. College of Mathematics and Physicsb. College of Data Science, Qingdao University of Science and Technology, Qingdao 266061, China)


Abstract: In this paperwe propose a drug-drug interactions (DDIs) prediction method called SA-BiLSTMwhich is based on bi-directional long short term memory (BiLSTM) and self-attention mechanism. Firstthe characteristic information of the drug is extracted using FP3 fingerprintsMACCS fingerprintsPubchem fingerprintsand PaDEL molecular descriptors. SecondLasso is applied to eliminate redundant features. Thenrepeated edited nearest neighbors (RENN) method is used to balance the data to obtain the best feature vector. Finallythe best feature vectors are imported into the classifier combined with self-attention and BiLSTM for predicting DDIs. SA-BiLSTM achieves high prediction accuracy on both datasets based on 5-fold cross-validation and comparison with other prediction methods. To further evaluate the predictive performance of SA-BiLSTMthe drug-drug interaction network is validated. The experimental results show that SA-BiLSTM achieves better prediction and can provide a new idea for predicting DDIs.


Key words: drug-drug interactions; feature extraction; RENN; BiLSTM; self-attention mechanism


收稿日期: 2023-11-18

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: 张明香(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)