Neurosymbolic multi-hop reasoning on knowledge graphs for drug discovery
Explainable Biomedical Recommendations via Reinforcement Learning Reasoning on Knowledge Graphs
A neurosymbolic approach of multi-hop reasoning on knowledge graphs has been shown to produce transparent explanations.
However, there is a lack of research applying it to complex biomedical datasets and problems to draw solid conclusions on its applicability.
In this paper, the approach is systematically applied to multiple biomedical datasets and recommendation tasks with fair benchmark comparisons to draw solid conclusions on its applicability for drug discovery.
The approach is found to outperform the best baselines by 21.7% on average whilst producing novel, biologically relevant explanations.
Gavin Edwards, Sebastian Nilsson, Benedek Rozemberczki, Eliseo Papa