Scholarlink

Polypharmacy, or the simultaneous use of multiple medications, increases the risk of Drug-Drug Interactions (DDIs), which can lead to adverse effects and reduced therapeutic efficacy. Traditional methods for identifying DDIs, such as clinical trials and post-market surveillance, are reactive and inefficient. To address this challenge, we developed a Graph Convolutional Network (GCN) model to predict DDIs using molecular and biological features. Our model processes molecular graphs, where drugs are represented by their atomic structures and bonds, enabling the GCN to capture complex molecular relationships. Key features, including molecular fingerprints, biological target profiles, and solubility data, were integrated into the model for comprehensive DDI prediction (DDIp). The model was trained on publicly available datasets and evaluated using metrics such as MSE, RMSE, MAE, and R-squared, demonstrating superior performance over traditional Machine Learning (ML) approaches. The study highlights the potential of GCNs in enhancing drug safety by predicting harmful interactions before they occur.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Molecular Feature-Based Prediction of Drug-Drug Interactions Using Graph Neural Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Flaviu-Ioan Gheorghita,
  • Dănut-Ovidiu Pop,
  • László Barna Iantovics

摘要

Polypharmacy, or the simultaneous use of multiple medications, increases the risk of Drug-Drug Interactions (DDIs), which can lead to adverse effects and reduced therapeutic efficacy. Traditional methods for identifying DDIs, such as clinical trials and post-market surveillance, are reactive and inefficient. To address this challenge, we developed a Graph Convolutional Network (GCN) model to predict DDIs using molecular and biological features. Our model processes molecular graphs, where drugs are represented by their atomic structures and bonds, enabling the GCN to capture complex molecular relationships. Key features, including molecular fingerprints, biological target profiles, and solubility data, were integrated into the model for comprehensive DDI prediction (DDIp). The model was trained on publicly available datasets and evaluated using metrics such as MSE, RMSE, MAE, and R-squared, demonstrating superior performance over traditional Machine Learning (ML) approaches. The study highlights the potential of GCNs in enhancing drug safety by predicting harmful interactions before they occur.