A survey on adverse drug reaction studies: data, tasks and machine learning methods

Area of Science:

• Pharmacology
• Computational Biology
• Data Science

Background:

• Adverse drug reaction (ADR) studies are vital for drug discovery and safety.
• The increasing volume of clinical and non-clinical data necessitates advanced analytical tools.
• Machine learning (ML) methods are increasingly employed for ADR analysis and prediction.

Purpose of the Study:

• To review and summarize existing ADR studies, focusing on ML applications.
• To categorize ADR studies into data creation, prediction, and mechanism analysis.
• To compare the performance of different ML methods in drug-ADR prediction.

Main Methods:

• Systematic review of ADR literature.
• Focus on ML techniques applied to ADR data.
• Comparative analysis of ML model performance for drug-ADR prediction.

Main Results:

• Summary of ADR data sources and relevant studies.
• Overview of ML methods applied across three key ADR tasks.
• Performance comparison of ML models for drug-ADR prediction.

Conclusions:

• ML shows significant promise in advancing ADR research.
• Challenges remain in data creation, prediction accuracy, and mechanism elucidation.
• Further research is needed to address open problems in ADR studies.