Abstract
Biomedical knowledge graphs encode domain knowledge as biomedical entities and relationships between them. Graph traversal algorithms can make use of these rich sources for the discovery of novel research hypotheses, e.g. the repurposing of a known drug. Traversed paths can serve to explain the underlying causal mechanisms. Most of these models, however, are trained to optimise for accuracy w.r.t. known gold standard drug-disease pairs, rather than for the explanatory mechanisms supporting such predictions. In this work, we aim to improve the retrieval of these explanatory mechanisms by improving path quality. We build on a reinforcement learning-based multi-hop reasoning approach for drug repurposing. First, we define a metric for path quality based on coherence with context entities. To calculate coherence, we learn a set of phenotype annotations with rule mining. Second, we use both the metric and the annotations to formulate a novel reward function. We assess the impact of contextual knowledge in a quantitative and qualitative evaluation, measuring: (i) the effect training with context has on the quality of reasoning paths, and (ii) the effect of using context for explainability purposes, measured in terms of plausibility, novelty, and relevancy. Results indicate that learning with contextual knowledge significantly increases path coherence, without affecting the interpretability for the domain experts.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
Initially, intervention nodes were extracted as well, but many of these proved too coarse grained—e.g. viral agents—to be useful for multi-hop reasoning.
- 11.
- 12.
- 13.
See the following concept for diabetes: https://data.cochrane.org/concepts/r4hp38bjj6qx, which they indicate is linked to unique codes from MedDRA: (10012594,10012601), MeSH: (D003920), and UMLS: (C0011849).
- 14.
- 15.
References
Agrawal, M., Zitnik, M., Leskovec, J.: Large-scale analysis of disease pathways in the human interactome. In: Pacific Symposium on Biocomputing, no. 212669, pp. 111–122 (2018). https://doi.org/10.1142/9789813235533_0011
Bahler, D., Stone, B., Wellington, C., Bristol, D.W.: Symbolic, neural, and Bayesian machine learning models for predicting carcinogenicity of chemical compounds. J. Chem. Inf. Comput. Sci. 40(4), 906–914 (2000). https://doi.org/10.1021/ci990116i
Bakal, G., Talari, P., Kakani, E.V., Kavuluru, R.: Exploiting semantic patterns over biomedical knowledge graphs for predicting treatment and causative relations. J. Biomed. Inform. 82(January), 189–199 (2018). https://doi.org/10.1016/j.jbi.2018.05.003
Blomqvist, E., Alirezaie, M., Santini, M.: Towards causal knowledge graphs-position paper. In: KDH@ ECAI (2020)
Brown, D.G., Wobst, H.J.: A decade of FDA-approved drugs (2010–2019): trends and future directions. J. Med. Chem. 64(5), 2312–2338 (2021). https://doi.org/10.1021/acs.jmedchem.0c01516
Das, R., et al.: Go for a walk and arrive at the answer: reasoning over paths in knowledge bases using reinforcement learning. In: 6th International Conference on Learning Representations, ICLR 2018 - Conference Track Proceedings (2018)
De Cao, N., Kipf, T.: MolGAN: an implicit generative model for small molecular graphs. arXiv preprint arXiv:1805.11973 (2018)
DiMasi, J.A., Grabowski, H.G., Hansen, R.W.: Innovation in the pharmaceutical industry: new estimates of R &D costs. J. Health Econ. 47, 20–33 (2016). https://doi.org/10.1016/j.jhealeco.2016.01.012
Drancé, M., Boudin, M., Mougin, F., Diallo, G.: Neuro-symbolic XAI for computational drug repurposing. In: KEOD, vol. 2, pp. 220–225 (2021)
Fu, G., Ding, Y., Seal, A., Chen, B., Sun, Y., Bolton, E.: Predicting drug target interactions using meta-path-based semantic network analysis. BMC Bioinform. 17(1), 1–10 (2016)
Galárraga, L., Teflioudi, C., Hose, K., Suchanek, F.M.: Fast rule mining in ontological knowledge bases with AMIE+. VLDB J. 24(6), 707–730 (2015). https://doi.org/10.1007/s00778-015-0394-1
Garijo, D., et al.: Towards automated hypothesis testing in neuroscience. In: Gadepally, V., et al. (eds.) DMAH/Poly -2019. LNCS, vol. 11721, pp. 249–257. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-33752-0_18
Gu, Y., Guan, Y., Missier, P.: Towards learning instantiated logical rules from knowledge graphs. arXiv preprint arXiv:2003.06071 (2020)
Guo, L., Sun, Z., Hu, W.: Learning to exploit long-term relational dependencies in knowledge graphs. In: International Conference on Machine Learning, pp. 2505–2514. PMLR (2019)
de Haan, R., Tiddi, I., Beek, W.: Discovering research hypotheses in social science using knowledge graph embeddings. In: Verborgh, R., et al. (eds.) ESWC 2021. LNCS, vol. 12731, pp. 477–494. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-77385-4_28
Himmelstein, D.S., et al.: Systematic integration of biomedical knowledge prioritizes drugs for repurposing. Elife 6, e26726 (2017)
Jaundoo, R., Craddock, T.J.: DRUGPATH: the drug gene pathway meta-database. Int. J. Mol. Sci. 21(9), 3171 (2020). https://doi.org/10.3390/ijms21093171
Jumper, J., et al.: Highly accurate protein structure prediction with AlphaFold. Nature 596(7873), 583–589 (2021)
Kundu, S.: AI in medicine must be explainable. Nat. Med. 27(8), 1328 (2021). https://doi.org/10.1038/s41591-021-01461-z
Lakkaraju, H., Kamar, E., Caruana, R., Leskovec, J.: Interpretable & explorable approximations of black box models. arXiv preprint arXiv:1707.01154 (2017)
Lin, X.V., Socher, R., Xiong, C.: Multi-hop knowledge graph reasoning with reward shaping. In: Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP 2018) (2018)
Liu, Y., Hildebrandt, M., Joblin, M., Ringsquandl, M., Raissouni, R., Tresp, V.: Neural multi-hop reasoning with logical rules on biomedical knowledge graphs. In: Verborgh, R., et al. (eds.) ESWC 2021. LNCS, vol. 12731, pp. 375–391. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-77385-4_22
Lv, X., et al.: Is multi-hop reasoning really explainable? Towards benchmarking reasoning interpretability. arXiv preprint arXiv:2104.06751 (2021)
Meilicke, C., Chekol, M.W., Ruffinelli, D., Stuckenschmidt, H.: Anytime bottom-up rule learning for knowledge graph completion. In: Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI). IJCAI/AAAI Press (2019)
Montavon, G., Samek, W., Müller, K.R.: Methods for interpreting and understanding deep neural networks. Digit. Sig. Process. 73, 1–15 (2018)
Nagarajan, M., et al.: Predicting future scientific discoveries based on a networked analysis of the past literature. In: Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 2019–2028 (2015)
Pankratius, V., et al.: Computer-aided discovery: toward scientific insight generation with machine support why scientists need machine support for discovery search. IEEE Intell. Syst. 31(4), 3–10 (2016). https://doi.org/10.1109/MIS.2016.60
Paul, D., Sanap, G., Shenoy, S., Kalyane, D., Kalia, K., Tekade, R.K.: Artificial intelligence in drug discovery and development. Drug Discov. Today 26(1), 80 (2021)
Percha, B., Altman, R.B.: A global network of biomedical relationships derived from text. Bioinformatics 34(15), 2614–2624 (2018)
Ribeiro, M.T., Singh, S., Guestrin, C.: “Why should i trust you?” Explaining the predictions of any classifier. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 1135–1144 (2016)
Saik, O.V., et al.: Novel candidate genes important for asthma and hypertension comorbidity revealed from associative gene networks. BMC Med. Genomics 11(1), 61–76 (2018)
Sosa, D.N., Derry, A., Guo, M., Wei, E., Brinton, C., Altman, R.B.: A literature-based knowledge graph embedding method for identifying drug repurposing opportunities in rare diseases. In: Pacific Symposium on Biocomputing 2020, pp. 463–474. World Scientific (2020)
Swanson, D.R., Smalheiser, N.R.: An interactive system for finding complementary literatures: a stimulus to scientific discovery. Artif. Intell. 91(2), 183–203 (1997). https://doi.org/10.1016/S0004-3702(97)00008-8
Tiddi, I., D’Aquin, M., Motta, E.: Walking linked data: a graph traversal approach to explain clusters. In: CEUR Workshop Proceedings (2014)
Wilcke, W.X., de Boer, V., de Kleijn, M.T., van Harmelen, F.A., Scholten, H.J.: User-centric pattern mining on knowledge graphs: an archaeological case study. J. Web Semant. 59, 100486 (2019). https://doi.org/10.1016/j.websem.2018.12.004
Williams, R.J.: Simple statistical gradient-following algorithms for connectionist reinforcement learning. Mach. Learn. 8(3), 229–256 (1992)
Zhou, X., Menche, J., Barabási, A.L., Sharma, A.: Human symptoms-disease network. Nat. Commun. 5(1), 1–10 (2014)
Acknowledgements
The authors would like to thankfully acknowledge Lotty Hooft and Alexandre Renaux for their valuable opinions and discussions related to this work. This work is supported by the EU Horizon 2020 research programme MUHAI, grant no. 951846.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Stork, L., Tiddi, I., Spijker, R., ten Teije, A. (2023). Explainable Drug Repurposing in Context via Deep Reinforcement Learning. In: Pesquita, C., et al. The Semantic Web. ESWC 2023. Lecture Notes in Computer Science, vol 13870. Springer, Cham. https://doi.org/10.1007/978-3-031-33455-9_1
Download citation
DOI: https://doi.org/10.1007/978-3-031-33455-9_1
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-33454-2
Online ISBN: 978-3-031-33455-9
eBook Packages: Computer ScienceComputer Science (R0)