Explainable Biomedical Recommendations via Reinforcement Learning Reasoning on Knowledge Graphs

For Artificial Intelligence to have a greater impact in biology and medicine,\nit is crucial that recommendations are both accurate and transparent. In other\ndomains, a neurosymbolic approach of multi-hop reasoning on knowledge graphs\nhas been shown to produce transparent explanations. However, there is a lack of\nresearch applying it to complex biomedical datasets and problems. In this\npaper, the approach is explored for drug discovery to draw solid conclusions on\nits applicability. For the first time, we systematically apply it to multiple\nbiomedical datasets and recommendation tasks with fair benchmark comparisons.\nThe approach is found to outperform the best baselines by 21.7% on average\nwhilst producing novel, biologically relevant explanations.\n