Machine learning to predict food effects during drug development: a comprehensive review

Area of Science:

• Pharmacokinetics and Drug Development
• Computational Biology and Bioinformatics

Background:

• Food consumption significantly impacts drug absorption, affecting drug efficacy and safety.
• Traditional in vitro and in vivo models struggle to accurately predict food effects (FE) due to gastrointestinal variability.

Purpose of the Study:

• To evaluate the predictive accuracy of machine learning (ML) models for food effects (FE) compared to conventional methods.
• To explore how ML utilizes food datasets to improve drug formulation and dosing strategies.

Main Methods:

• Review and analysis of existing literature on ML applications in predicting food effects (FE).
• Examination of supervised, unsupervised, and reinforcement learning techniques for absorption, distribution, metabolism, and elimination (ADME) forecasting.
• Discussion of hybrid approaches combining ML with physiologically based pharmacokinetic (PBPK) modeling.

Main Results:

• Machine learning (ML) demonstrates potential in predicting food effects (FE), offering advantages over traditional methods.
• ML models can leverage food dataset information to optimize drug formulation and dosing.
• Challenges remain regarding data quality, model generalizability, and integration into drug development pipelines.

Conclusions:

• Machine learning (ML) shows significant promise for predicting food effects (FE) in drug development.
• Integrating ML with PBPK modeling and addressing current challenges can enhance its utility.
• Interdisciplinary approaches incorporating explainable AI and ethical considerations are crucial for advancing ML in pharmacokinetics and patient care.