Predictive Modeling for Drug Repurposing Using Side-Effect and Toxicity Similarities

Key Points

• The aim is to develop a predictive model for drug repurposing based on side-effect and toxicity similarities.
• Utilized databases of toxicity assessments from in-vitro testing of thousands of chemicals.
• Developed a machine learning model to predict shared protein targets based on structural, side-effect, and toxicity similarities.
• Validated the model using the top 10 drug-drug pairs and uncharted chemicals to assess its predictive capability.
• The predictive model identified promising drug candidates efficiently, demonstrating robustness in identifying shared targets.
• Achieved performance comparable to other non-linear models while offering a simpler and interpretable logistic regression approach.