A Novel Evaluation Method and Learning Approach for Identifying and Addressing Interaction Type Recognition Issues in Drug-Drug Interactions Prediction

Key Points

• The aim is to improve the recognition of interaction types in drug-drug interactions and address existing limitations in prediction methods.
• Introduced a novel interaction type recognition test to evaluate model performance.
• Developed a learning approach incorporating negative pairs to enhance interaction type learning.
• Conducted experiments on two structurally different datasets: Drugbank and Twosides.
• Both datasets showed significant improvement in interaction type recognition performance.
• The proposed method effectively enhances the model's learning of interaction types.