Novel Customized Semantic Representation and Machine Learning for Accurate Forecasting of Drug-Drug Interaction Severity

This project focuses on the accurate forecasting of Drug-Drug Interaction (DDI) severity using machine learning models. The dataset contains various classes, such as "mechanism_similarity," "target_overlap," "side_effect_overlap," and "enzyme_inhibition_score," among others. These classes represent different aspects of drug interactions, such as molecular similarity, bioavailability, and clinical evidence. To achieve precise predictions, we employed machine learning models like Random Forest and XGBoost, which were trained to predict the seriousness of drug interactions based on these features. The target variable, "seriousness_label," includes classes such as "contraindicated," "minor," "major," and "moderate," representing different levels of severity in drug interactions. The models were optimized to ensure high performance in classifying drug interactions accurately, aiding in the identification of potential risks in pharmaceutical treatment. This solution has significant implications for drug safety and pharmacovigilance, helping healthcare professionals assess the severity of interactions and make informed decisions.

Keywords: Drug-Drug Interaction, Random Forest, XGBoost, Machine Learning, Drug Safety, Pharmacovigilance, Seriousness Prediction, Pharmaceutical Risk Management.