MTrans: M-Transformer and Knowledge Graph-Based Network for Predicting Drug–Drug Interactions

Author:

Wu Shiqi1ORCID,Liu Baisong1ORCID,Zhang Xueyuan1,Shao Xiaowen1,Lin Chennan1

Affiliation:

1. Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo 315000, China

Abstract

The combined use of multiple medications is common in treatment, which may lead to severe drug–drug interactions (DDIs). Deep learning methods have been widely used to predict DDIs in recent years. However, current models need help to fully understand the characteristics of drugs and the relationships between these characteristics, resulting in inaccurate and inefficient feature representations. Beyond that, existing studies predominantly focus on analyzing a single DDIs, failing to explore multiple similar DDIs simultaneously, thus limiting the discovery of common mechanisms underlying DDIs. To address these limitations, this research proposes a method based on M-Transformer and knowledge graph for predicting DDIs, comprising a dual-pathway approach and neural network. In the first pathway, we leverage the interpretability of the transformer to capture the intricate relationships between drug features using the multi-head attention mechanism, identifying and discarding redundant information to obtain a more refined and information-dense drug representation. However, due to the potential difficulty for a single transformer model to understand features from multiple semantic spaces, we adopted M-Transformer to understand the structural and pharmacological information of the drug as well as the connections between them. In the second pathway, we constructed a drug–drug interaction knowledge graph (DDIKG) using drug representation vectors obtained from M-Transformer as nodes and DDI types as edges. Subsequently, drug edges with similar interactions were aggregated using a graph neural network (GNN). This facilitates the exploration and extraction of shared mechanisms underlying drug–drug interactions. Extensive experiments demonstrate that our MTrans model accurately predicts DDIs and outperforms state-of-the-art models.

Funder

Natural Science Foundation of Zhejiang Province

Science and Technology Innovation 2025 Major Project of Ningbo

Natural Science Foundation of Ningbo

Research and Development of a Digital Infrastructure Cloud Operation and Maintenance Platform Based on 5G and AI

China Innovation Challenge (Ningbo) Major Project

Publisher

MDPI AG

Reference41 articles.

1. “Health for all” and the challenges for pharmaceutical policies: A critical interpretive synthesis over 40 years;Gautier;Soc. Sci. Humanit. Open,2022

2. In silico methods to address polypharmacology: Current status, applications and future perspectives;Lavecchia;Drug Discov. Today,2016

3. A reference set of clinically relevant adverse drug-drug interactions;Kontsioti;Sci. Data,2022

4. Adverse drug reactions: Definitions, diagnosis, and management;Edwards;Lancet,2000

5. Trends in Prescription Drug Use Among Adults in the United States from 1999–2012;Kantor;JAMA,2015

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3