PREDICTING DRUG - DRUG INTERACTIONS BASED ON INTEGRATED SIMILARITY AND SEMI SUPERVISED LEARNING

Abstract

Drug-drug interactions (DDIs) are a major concern in healthcare, as they can lead to adverse drug reactions, reduced therapeutic effectiveness, or even life-threatening conditions. With the increasing use of multi-drug therapies, there is a critical need for efficient methods to predict potential DDIs. In this paper, we propose a novel computational framework named DDI-IS-SL, which integrates multiple drug similarity measures with a semi-supervised learning approach. The model combines chemical, biological, and phenotypic data to compute similarity using cosine similarity and Gaussian Interaction Profile (GIP) kernel methods. A Regularized Least Squares (RLS) classifier is employed to predict interaction probabilities between drug pairs. Experimental results using cross-validation techniques demonstrate that the proposed model achieves high accuracy and outperforms existing methods. The system is scalable, efficient, and capable of predicting interactions for new drugs, making it suitable for real-world healthcare applications.