Boonyarit Bundit, Kositchutima Matin, Phattalung Tisorn Na, Yamprasert Nattawin, Thuwajit Chanitra, Rungrotmongkol Thanyada, Nutanong Sarana
School of Information Science and Technology, Vidyasirimedhi Institute of Science and Technology, Rayong 21210, Thailand.
Kamnoetvidya Science Academy, Rayong 21210, Thailand.
Gigascience. 2025 Jan 6;14. doi: 10.1093/gigascience/giaf080.
Drug combination therapy plays a pivotal role in addressing the molecular heterogeneity of cancer, improving treatment efficacy, minimizing resistance, and reducing toxicity. Deep learning approaches have significantly advanced drug combination discovery by addressing the limitations of conventional laboratory experiments, which are time-consuming and costly. While most existing models rely on the molecular structure of drugs and gene expression data, incorporating protein-level expression provides a more accurate representation of cellular behavior and drug responses. In this study, we introduce SynProtX, an enhanced deep learning model that explicitly integrates large-scale proteomics with deep neural networks (DNNs) and the molecular structure of drugs with graph neural networks (GNNs).
The SynProtX-GATFP model, which combines molecular graphs and fingerprints through a graph attention network architecture, demonstrated superior predictive performance for the FRIEDMAN study dataset. We further evaluated its cell line-specific performance, which achieved accuracy across diverse tissue and study datasets. By incorporating protein expression data, the model consistently enhanced predictive performance over gene expression-only models, reflecting the functional state of cancer cells. The generalizability of SynProtX was rigorously validated using cold-start prediction, including leave-drug-combination-out, leave-drug-out, and leave-cell-line-out validation strategies, highlighting its robust performance and potential for clinical applicability. Additionally, SynProtX identified key cancer-associated proteins and molecular substructures, offering novel insights into the biological mechanisms underlying drug synergy. These findings highlight the potential of integrating large-scale proteomics and multiomics data to advance anticancer drug design and combination therapy strategies for personalized medicine. Availability and implementation: https://github.com/manbaritone/SynProtX.
联合药物疗法在应对癌症的分子异质性、提高治疗效果、最小化耐药性以及降低毒性方面发挥着关键作用。深度学习方法通过克服传统实验室实验耗时且成本高的局限性，显著推动了联合药物的发现。虽然大多数现有模型依赖于药物的分子结构和基因表达数据，但纳入蛋白质水平的表达能更准确地反映细胞行为和药物反应。在本研究中，我们引入了SynProtX，这是一种增强的深度学习模型，它将大规模蛋白质组学与深度神经网络（DNN）以及药物的分子结构与图神经网络（GNN）进行了明确整合。
通过图注意力网络架构将分子图和指纹相结合的SynProtX - GATFP模型，在FRIEDMAN研究数据集上展现出卓越的预测性能。我们进一步评估了其细胞系特异性性能，该模型在不同组织和研究数据集上均实现了较高的准确率。通过纳入蛋白质表达数据，该模型相对于仅使用基因表达的模型持续提升了预测性能，反映了癌细胞的功能状态。使用冷启动预测（包括留用联合药物、留用药物和留用细胞系验证策略）对SynProtX的泛化能力进行了严格验证，突出了其强大的性能和临床应用潜力。此外，SynProtX识别出了关键的癌症相关蛋白质和分子亚结构，为药物协同作用的生物学机制提供了新的见解。这些发现凸显了整合大规模蛋白质组学和多组学数据以推进抗癌药物设计和个性化医学联合治疗策略的潜力。可用性和实现方式：https://github.com/manbaritone/SynProtX 。
