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

Area of Science:

• Computational biology
• Pharmacology
• Artificial Intelligence

Background:

• Drug repositioning offers a cost-effective alternative to de novo drug discovery, especially for diseases with unmet needs.
• Machine Learning (ML) and Deep Learning (DL) are increasingly vital for analyzing large datasets to identify novel drug indications efficiently.
• These AI approaches enhance the speed, accuracy, and efficiency of discovering new therapeutic uses for existing medications.

Purpose of the Study:

• To systematically review recent advancements in ML and DL for drug repositioning.
• To analyze model architectures, methodologies, data types, and evaluation metrics in AI-driven drug repurposing.
• To identify key findings and trends in the application of AI for identifying novel drug indications.

Main Methods:

• Systematic review of 24 studies published between 2015 and 2025.
• Analysis of model architectures (e.g., GCN, DNN, RF, SVM), methodologies, and evaluation metrics.
• Investigation of data types and benchmark databases (e.g., DrugBank, PubChem, ChEMBL) used in drug repositioning studies.

Main Results:

• Deep learning (GCN, DNN) and machine learning (RF, SVM) models are prevalent in drug repositioning.
• Predictive models demonstrate high performance, with accuracy often exceeding 90% in relevant studies.
• Drug-Target Interaction (DTI) prediction frequently utilizes benchmark databases like DrugBank, PubChem, and ChEMBL.

Conclusions:

• Advanced deep learning frameworks are becoming central to modern drug repurposing efforts.
• A significant translational gap exists between in silico predictions and clinical outcomes due to interpretability and data integrity challenges.
• Future research should focus on interpretable AI, advanced DL, and hybrid models to enhance clinical translation, particularly for rare diseases.