cs.LGMar 13, 2026

Enhanced Drug-drug Interaction Prediction Using Adaptive Knowledge Integration

Authors: Pengfei Liu, Jun Tao, Zhixiang Ren

Organizations: Pengcheng Laboratory, Shenzhen 518000 · School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006, China

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Drug-drug interaction event (DDIE) prediction is crucial for preventing adverse reactions and ensuring optimal therapeutic outcomes. However, existing methods often face challenges with imbalanced datasets, complex interaction mechanisms, and poor generalization to unknown drug combinations. To address these challenges, we propose a knowledge augmentation framework that adaptively infuses prior drug knowledge into a large language model (LLM). This framework utilizes reinforcement learning techniques to facilitate adaptive knowledge extraction and synthesis, thereby efficiently optimizing the strategy space to enhance the accuracy of LLMs for DDIE predictions. As a result of few-shot learning, we achieved a notable improvement compared to the baseline. This approach establishes an effective framework for scientific knowledge learning for DDIE predictions.

Figures & tables

 
Table 1
Fig. 1 : Overview of the framework. Leveraging the information of drug molecules to enhance the accuracy of DDIE prediction in few-shot scenarios. (a) Knowledge Extraction classifies drug types from features using specific clustering methods and counts. (b) Prompt Manager uses drug types as prior knowledge and combines them with molecular representations or descriptions to form prompts. (c) LLM-based DDIE Predictor predicts DDIEs, with the performance feedback serving as a reward to the Adaptive Strategy Selector. (d) Adaptive Strategy Selector develops strategies based on clustering methods, knowledge synthesis strategies, and LLM hyperparameters, exploring the strategy space to iteratively select optimal states.
Fig. 2 : Dataset analysis. (a) and (b) show the frequency distribution of DDIE categories in the DeepDDI 2 and DDIMDL datasets. Categories are grouped into Common, Few, and Rare based on sample frequency, with the y-axis showing the number of samples and the x-axis showing the category indices. Note that although some numerical labels may overlap across datasets, the underlying categories differ.
DatasetFrequencyTrainValidTest
DeepDDI 2 [ 17 ]Common (50 << )44,12644,113132,110
 Few (15-50)108128298
 Rare ( << 15)433485
DDIMDL [ 44 ]Common (50 << )7,3317,30821,958
 Few (15-50)7686263
 Rare ( << 15)203362
Table 1: Distribution of data samples across different interaction frequency categories in training, validation, and test sets.
Fig. 3 : Accuracy performance of baseline models and ours on DeepDDI 2 (a) and DDIMDL (b) datasets across data splits (All Events, Common, Few, Rare).
Fig. 4 : The exploration process of strategy combinations for the few and rare data splits in DeepDDI 2 dataset. The horizontal axis represents the exploration steps, while the vertical axis shows the F1-score evaluation metric. Different colors indicate the corresponding clustering methods. The top-3 F1 scores in each split are highlighted with red circles. The right axes show the corresponding Reward (purple dashed line) and Validation Loss (crimson dash-dot line).
Fig. 5 : Ablation study on DeepDDI 2 (a) and DDIMDL (b) datasets for Few and Rare splits. “GR” denotes coarse grid search, “RS” denotes random search, and “Ours” is the proposed RL-based searcher. Error bars indicate standard deviation over three runs.
Fig. 6 : Cluster analysis and case studies. (a) Agglomerative clustering consistency analysis. It compares the consistency of agglomerative clustering with 6 and 8 clusters, visualizing the clustering results of samples split into clusters 0 and 2 in the 6-cluster analysis to observe distribution differences. (b) Cross-classification analysis of class B drugs and case studies. The samples from clusters 0 and 2 in the 6-cluster analysis were selected, focusing on drugs from DrugBank’s BLOOD AND BLOOD FORMING ORGANS category (B) to observe drug class distribution. The ten molecules from cluster 2 in the 6-cluster analysis were chosen as case studies to analyze the basis and validity of the classification.

Explore similar work

CardsList

  1. DDIAgents: Mechanism-Conditioned Context Flow for Drug-Drug Interaction Prediction

    Jun 30, 2026Zhenqian Shen, Yu Liu, Xiaoyi Fu +1Multi-Agent LLM SystemsAI for Science

    Department of Electronic Engineering, Tsinghua University Beijing, China · Institute of Biomedical Engineering, University of Oxford Oxford, United Kingdom · Division of Emerging Interdisciplinary Areas, Hong Kong University of Science and Technology Hong Kong, China

  2. ColdDDI: Evaluating Knowledge Utilization in Cold-Start Drug-Drug Interaction Prediction

    Oct 4, 2026Jiheng Liang, Chen Zhao, Di Wu +4Biomedical KGs

    Data Science, William & Mary · Computer Science, Baylor University · Southwest University +2

  3. CoFEND: A Cross-Modal Fusion End-to-End Network for Cold-Start Drug-Drug Interaction Prediction

    Jul 3, 2026Di Wu, Hongyi Sun, Haichao Xu +3Multimodal Graph LearningDrug Discovery

    School of Physics and Electronic Science, Zunyi Normal University