DVMPDC: A Deep Learning Model Based on Dual-View Representation and Multi-Strategy Pooling for Predicting Synergistic Drug Combinations

The development of drug resistance in cancer significantly affects treatment efficacy. Combination therapy holds promise in overcoming drug resistance and providing more effective treatment. However, discovering effective drug combinations using traditional experimental methods is time-consuming and labor-intensive. In recent years, a series of computational models have been developed to predict feasible drug combinations and improve efficiency. Nevertheless, most methods treat drugs in combinations as independent entities, neglecting the interaction information between drugs. Furthermore, many methods adopt a single strategy to evaluate the importance of atoms when obtaining graph embeddings from atom feature matrices, failing to fully consider the complexity and diversity of graph structures. In this study, we propose a deep learning framework for collaborative drug combination prediction, named DVMPDC. DVMPDC utilizes a multi-layer dual-view representation learning module to acquire hierarchical atomic interaction information and employs a multiple strategy-based attention pooling operator to obtain graph embeddings. Experimental results demonstrate that DVMPDC outperforms existing state-of-the-art models on all performance metrics. The results of the substructure visualization experiments demonstrate the significant potential of our model in capturing important substructures of drugs. Additionally, some case studies also validate the effectiveness of our model in practical applications. The codes of DVMPDC are available at https://github.com/xcl272/DVMPDC .