Parameterized SVM for personalized drug concentration prediction.

Area of Science:

• Pharmacometrics
• Machine Learning in Healthcare
• Computational Biology

Background:

• Accurate modeling of Drug Concentration to Time (DCT) curves is crucial for personalized medicine.
• Existing methods may not fully leverage patient-specific features or adapt to real-world measurements.
• Therapeutic Drug Monitoring (TDM) provides valuable data for refining pharmacokinetic models.

Purpose of the Study:

• To propose a novel Parameterized Support Vector Machine (ParaSVM) for robust DCT curve modeling.
• To integrate patient features and analytical modeling for improved DCT curve approximation.
• To enable adaptive calibration of DCT curves using TDM data.

Main Methods:

• Development of the Parameterized Support Vector Machine (ParaSVM) algorithm.
• Integration of Support Vector Machine (SVM) for patient feature analysis.
• Application of an analytical model for DCT point approximation and curve calibration.
• Utilization of the RANSAC algorithm to build a parameter library for basis functions.

Main Results:

• Successful implementation of ParaSVM for constructing DCT curves.
• Demonstrated calibration of DCT curves using TDM measurements in an imatinib case study.
• Validation of the ParaSVM approach's ability to model and refine DCT profiles.

Conclusions:

• ParaSVM offers a powerful and flexible framework for DCT curve modeling.
• The integration of SVM and analytical modeling enhances predictive accuracy.
• ParaSVM facilitates effective curve calibration with real-world TDM data, improving therapeutic drug management.