Multi-Knowledge Graph and Multi-View Entity Feature Learning for Predicting Drug-Related Side Effects DOI
Ping Xuan, Tianhong Cheng, Hui Cui

et al.

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: May 6, 2025

Computational prediction of potential drug side effects plays a crucial role in reducing health risks for clinical patients and accelerating development. Recent methods have constructed heterogeneous graphs that represent drugs their effects, utilizing graph learning strategies such as convolutional networks to predict associations between them. However, existing approaches fail fully exploit the diverse topologies semantics present multiple knowledge graphs. We propose MVDSA, novel multi-view drug-side effect association model. Our approach integrates relationship semantics, local graphs, features entity pairs. First, we two based on functional structural similarity, associations. These capture topological semantic connections entities from perspectives. Second, considering similarities entities, designed space-sensitive strategy where relation-gated encoder is each type relationship. This adaptively adjusts contribution feature relational representation, facilitating entity-specific within space. Third, given types head tail connection-sensitive attention mechanism integrate these relationships. To different learning, graph-level fuse enhanced Finally, multi-layer perceptron (MLP) encode pairs three perspectives entities. Extensive experiments demonstrate MVDSA outperforms 10 state-of-the-art predicting Ablation studies validate contributions proposed innovations improved performance. Additionally, case candidate five highlight MVDSA's capability discover

Language: Английский

Multi-Knowledge Graph and Multi-View Entity Feature Learning for Predicting Drug-Related Side Effects DOI
Ping Xuan, Tianhong Cheng, Hui Cui

et al.

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: May 6, 2025

Computational prediction of potential drug side effects plays a crucial role in reducing health risks for clinical patients and accelerating development. Recent methods have constructed heterogeneous graphs that represent drugs their effects, utilizing graph learning strategies such as convolutional networks to predict associations between them. However, existing approaches fail fully exploit the diverse topologies semantics present multiple knowledge graphs. We propose MVDSA, novel multi-view drug-side effect association model. Our approach integrates relationship semantics, local graphs, features entity pairs. First, we two based on functional structural similarity, associations. These capture topological semantic connections entities from perspectives. Second, considering similarities entities, designed space-sensitive strategy where relation-gated encoder is each type relationship. This adaptively adjusts contribution feature relational representation, facilitating entity-specific within space. Third, given types head tail connection-sensitive attention mechanism integrate these relationships. To different learning, graph-level fuse enhanced Finally, multi-layer perceptron (MLP) encode pairs three perspectives entities. Extensive experiments demonstrate MVDSA outperforms 10 state-of-the-art predicting Ablation studies validate contributions proposed innovations improved performance. Additionally, case candidate five highlight MVDSA's capability discover

Language: Английский

Citations

0