Machine Learning for Prediction of Drug Concentrations: Application and Challenges

Area of Science:

• Pharmacokinetics
• Machine Learning
• Computational Biology

Background:

• Machine learning (ML) is increasingly applied to pharmacokinetics (PK) analysis due to algorithmic advancements and data availability.
• Pharmacokinetic studies benefit from ML for predicting drug concentrations and understanding drug disposition.

Purpose of the Study:

• To review machine learning applications in pharmacokinetics up to September 2024.
• To summarize ML algorithms, data preprocessing, application scenarios, and challenges in PK analysis.
• To evaluate ML model performance against traditional population pharmacokinetics (PopPK) models.

Main Methods:

• Systematic literature search of PubMed and IEEE Xplore databases.
• Analysis of studies focusing on ML for drug concentration prediction in pharmacokinetics.
• Categorization of ML algorithms, data preprocessing techniques, and application areas.

Main Results:

• Tree-based algorithms and neural networks are the most frequently used ML methods in PK.
• ML models demonstrate performance comparable to conventional PopPK models.
• Ensemble modeling, particularly combining ML and pharmacometrics, improves prediction accuracy.

Conclusions:

• Machine learning offers a powerful alternative for pharmacokinetics analysis, especially for drug concentration prediction.
• Continued research into ML algorithms and ensemble techniques is crucial for advancing PK modeling.
• Addressing current challenges will further integrate ML into routine pharmacokinetic assessments.