A Machine Learning Framework for Drug-Drug Interaction Prediction Using Smiles Features

Predicting drug-drug interactions in disease treat- ments is crucial. Identifying these interactions can significantly impact the effectiveness and safety of a treatment. Efforts have been focused on utilizing knowledge about drug-drug interactions. However, not all interactions are known. This study presents machine learning-based methods - KNN, logistic regression, SVM, random forest, XGBoost, and a convolution neural network - to predict drug-drug interactions. We improved the dataset using the DrugBank database and proposed a learning framework for efficient prediction. The results demonstrate that the proposed framework performs well, especially when using SVM and logistic regression as the learning engines.