上表展示了我们的实验结果,可以看到我们在除 DDI 率之外的所有指标上,均取得了 SOTA 的表现。至于 DDI 率,我们在一众方法里表现第二好,排名第一的是 4SDrug [9]。4SDrug 在 DDI 率最低得益于它倾向于推荐较少的药物,这使得它在别的指标上表现不如意。此外,我们还做了一些 ablation study 和 case study 进一步突出我们方法的有效性,这一部分实验详见原论文。
此外,我们的方法 MoleRec 已经作为一个 baseline model 被整合进 UIUC 的 Jimeng Sun 老师组发布的 PyHealth Package 中。截止目前,该项目已在GitHub上收获 star。
写在最后We're always open for possible collaborations and feel free to contact us.欢迎一起交流与进步!
参考文献
[1] Klekota, Justin, and Frederick P. Roth. "Chemical substructures that enrich for biological activity." Bioinformatics 24.21 (2008): 2518-2525.[2] Phanus-Umporn, Chuleeporn, et al. "Privileged substructures for anti-sickling activity via cheminformatic analysis." RSC advances 8.11 (2018): 5920-5935.[3] Yang, Nianzu, et al. "Learning substructure invariance for out-of-distribution molecular representations." Advances in Neural Information Processing Systems 35 (2022): 12964-12978.[4] Shang, Junyuan, et al. "Gamenet: Graph augmented memory networks for recommending medication combination." proceedings of the AAAI Conference on Artificial Intelligence. Vol. 33. No. 01. 2019.[5] Yang, Chaoqi, et al. "SafeDrug: Dual Molecular Graph Encoders for Recommending Effective and Safe Drug Combinations." 30th International Joint Conference on Artificial Intelligence, IJCAI 2021. International Joint Conferences on Artificial Intelligence, 2021.[6] Xu, Keyulu, et al. "How Powerful are Graph Neural Networks?." International Conference on Learning Representations. 2018.[7] Degen, Jörg, et al. "On the Art of Compiling and Using'Drug‐Like'Chemical Fragment Spaces." ChemMedChem: Chemistry Enabling Drug Discovery 3.10 (2008): 1503-1507.[8] Lee, Juho, et al. "Set transformer: A framework for attention-based permutation-invariant neural networks." International conference on machine learning. PMLR, 2019.[9] Tan, Yanchao, et al. "4sdrug: Symptom-based set-to-set small and safe drug recommendation." Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 2022.