The evolution of nonlinear mixed effects modeling in pharmacometrics: toward AI-based variational autoencoders

Area of Science:

• Pharmacometrics
• Computational Biology
• Statistical Modeling

Background:

• Nonlinear mixed effects (NLME) modeling has advanced significantly due to innovations in numerical methods and computing.
• Early NLME approaches relied on linearization, followed by sampling-based methods.

Purpose of the Study:

• To outline the historical evolution of nonlinear mixed effects (NLME) modeling.
• To highlight the emergence of artificial intelligence (AI) in shaping the future of NLME.
• To introduce AI-augmented pharmacometric (PMX) models.

Main Methods:

• Review of historical NLME modeling techniques, including linearization and sampling-based methods.
• Introduction of variational autoencoders as a bridge between classical NLME and AI.
• Conceptual framework for AI-augmented PMX models.

Main Results:

• NLME modeling is entering a new phase driven by AI.
• Variational autoencoders facilitate the integration of AI into NLME.
• AI-augmented PMX models offer enhanced capabilities for data integration and complex problem-solving.

Conclusions:

• The integration of AI, particularly through variational autoencoders, represents a significant advancement in NLME modeling.
• AI-augmented NLME models are poised to handle multimodal data and address more complex scientific challenges.
• This evolution promises to expand the scope and applicability of NLME in various scientific domains.