Machine Learning Techniques for Predicting Drug-Related Side Effects: A Scoping Review.

Area of Science:

• Pharmacovigilance and Computational Chemistry
• Drug Discovery and Development

Background:

• Accurate prediction of drug side effects is crucial for patient safety.
• Machine learning (ML) offers advanced methods for predicting adverse drug reactions.
• This review focuses on ML approaches utilizing chemical, biological, and phenotypical features.

Purpose of the Study:

• To review machine learning approaches for predicting drug-related side effects.
• To identify key features and algorithms used in drug safety prediction.
• To assess the potential of ML in improving drug development.

Main Methods:

• Scoping review methodology.
• Comprehensive literature search across multiple databases.
• Timeframe: January 1, 2013, to December 31, 2023.

Main Results:

• Random Forest, k-nearest neighbor, and support vector machine algorithms are widely used.
• Ensemble methods, especially Random Forest, highlight the importance of integrating chemical and biological features.
• Combining diverse features significantly enhances prediction accuracy for drug side effects.

Conclusions:

• A variety of features, datasets, and ML algorithms are essential for effective side effect prediction.
• Ensemble methods and Random Forest demonstrate superior performance.
• Integrating chemical and biological features improves prediction accuracy.
• ML holds significant potential for advancing drug development and clinical trials.
• Future research should explore specific feature types, selection methods, and graph-based approaches.