Drug Repurposing Using Machine Learning and Deep Learning: A Systematic Literature Review

Drug repositioning identifies new therapeutic applications for existing drugs, particularly for diseases lacking effective treatments. This approach can significantly reduce the time and cost of drug discovery. Recently, Machine Learning (ML) and Deep Learning (DL) have become pivotal tools in this field, facilitating the analysis of large-scale datasets and enhancing the accuracy, efficiency, and speed of identifying novel drug indications.

This systematic review examines recent advances in machine learning and deep learning approaches for drug repositioning. A comprehensive search of multiple databases yielded 24 relevant studies published between 2015 and 2025. The review analyzes model architectures, methodologies, and evaluation metrics, while also investigating data types and key findings related to the repositioned drugs.

Results show that deep learning architectures, such as Graph Convolutional Networks (GCN), Deep Neural Networks (DNN), alongside machine learning models like Random Forest (RF), and Support Vector Machines (SVM), are highly prevalent. Across all reviewed studies, predictive models consistently demonstrate strong performance, with most accuracy-based studies reporting values above 90%, while studies using other evaluation metrics also show competitive results. DrugBank, PubChem, and ChEMBL emerged as the primary benchmarked databases, particularly in studies aimed at predicting Drug-Target Interactions (DTIs). While research remains concentrated on oncology and neurodegenerative diseases, the field is rapidly expanding toward other conditions.

Our analysis demonstrates a clear move toward advanced deep learning frameworks, which have become the modern pillars for drug repurposing. However, a critical translational gap remains between successful in silico predictions and actual clinical outcomes, primarily due to challenges in model interpretability and data integrity.

AI-driven drug repositioning offers a cost-effective strategy to accelerate drug discovery. To achieve real-world medical impact, future research must prioritize interpretable architectures, advanced deep learning frameworks, and robust hybrid models, particularly for treating rare diseases, to bridge the gap between computational success and clinical implementation effectively.