Case-Based Reasoning Enhances the Predictive Power of LLMs in Drug-Drug Interaction

TL;DR

This work tackles drug-drug interaction prediction by marrying case-based reasoning with large language models. It builds a knowledge repository of pharmacological cases, augments LLM reasoning with GNN-derived drug associations, and uses a hybrid retrieval plus dual-layer prompting scheme to infer interaction types and mechanisms. Through representative sampling, it maintains a compact, diverse repository while achieving state-of-the-art results on DrugBank and TWOSIDES with strong interpretability. The approach holds practical potential for safer pharmacotherapy and provides a flexible, scalable framework for integrating pharmacological principles into LLM-driven reasoning.

Abstract

Drug-drug interaction (DDI) prediction is critical for treatment safety. While large language models (LLMs) show promise in pharmaceutical tasks, their effectiveness in DDI prediction remains challenging. Inspired by the well-established clinical practice where physicians routinely reference similar historical cases to guide their decisions through case-based reasoning (CBR), we propose CBR-DDI, a novel framework that distills pharmacological principles from historical cases to improve LLM reasoning for DDI tasks. CBR-DDI constructs a knowledge repository by leveraging LLMs to extract pharmacological insights and graph neural networks (GNNs) to model drug associations. A hybrid retrieval mechanism and dual-layer knowledge-enhanced prompting allow LLMs to effectively retrieve and reuse relevant cases. We further introduce a representative sampling strategy for dynamic case refinement. Extensive experiments demonstrate that CBR-DDI achieves state-of-the-art performance, with a significant 28.7% accuracy improvement over both popular LLMs and CBR baseline, while maintaining high interpretability and flexibility.

Figures (10)

• Figure 1: (a). Illustration of using historical cases to solve new cases in DDI task. (b). Accuracy comparison on DrugBank dataset: our CBR-DDI shows significant improvement over base model and Naive-CBR.
• Figure 2: Comparison between Naive-CBR method and our method CBR-DDI. CBR-DDI constructs a knowledge repository storing cases with rich pharmacological insights, and enhances LLM predictions via LLM-GNN collaborative case retrieval, dual-layer knowledge-enhanced reuse, and representative sampling-based dynamic refinement.
• Figure 3: Example from the knowledge repository.
• Figure 4: Impact of hybrid retriever's hyperparameter.
• Figure 5: One case study from DrugBank.