Dynamic Knowledge-Prior for Few-Shot Drug-Drug Interaction Events Prediction
Pengfei Liu, Jun Tao, Zhixiang Ren · 2024
Predicting drug-drug interaction events (DDIE) is crucial to minimizing harmful side effects and improving therapeutic results. However, existing approaches face challenges due to unbalanced datasets, the complex nature of interaction mechanisms, and their limited ability to apply to new drug combinations. To tackle these limitations, we introduce an innovative framework that adaptively infuses prior drug knowledge into a large language model (LLM). Our method utilizes reinforcement learning (RL) to optimize knowledge extraction and integration. This strategic use of RL reduces the complexity of exploration and hyperparameter tuning, leading to a significant enhancement in the prediction accuracy of the LLM for DDIEs. This framework not only enhances DDIE prediction but also paves the way for a more integrated approach to scientific knowledge in this context, potentially revolutionizing the field of pharmaceutical research.