Reduction of quantitative systems pharmacology models using artificial neural networks

Area of Science:

• Pharmacology
• Computational Biology
• Systems Biology

Background:

• Quantitative systems pharmacology (QSP) models are crucial for understanding biological systems but often exhibit high complexity, hindering simulation and analysis.
• Model-order reduction (MOR) offers a solution by creating simplified, mechanistically sound models from complex systems.
• Challenges arise in MOR for models with high-dimensional, discontinuous interfaces, such as those found in blood coagulation.

Purpose of the Study:

• To explore the application of artificial neural networks (ANNs) for approximating input-output relationships in high-dimensional QSP models.
• To demonstrate the efficacy of ANNs in model reduction for complex biological systems with challenging interfaces.
• To develop an efficient approximation method for complex models applicable to simulation and control.

Main Methods:

• Utilized artificial neural networks to approximate the input-output relationship of a complex, high-dimensional systems model.
• Applied the ANN approach to a specific model of blood coagulation featuring a discontinuous interface.
• Evaluated the accuracy and efficiency of the ANN-based approximation against the original complex model.

Main Results:

• The ANN approach successfully approximated the input-output relationship of the complex blood coagulation model.
• The method provided an efficient approximation with a desired level of accuracy.
• The technique demonstrated applicability to various complex models, offering significant speed improvements for simulation and control.

Conclusions:

• Artificial neural networks offer a viable and efficient method for model-order reduction in complex, high-dimensional systems pharmacology models.
• This approach overcomes challenges posed by discontinuous interfaces, enabling accurate approximations.
• The technique enhances the utility of complex models for simulation, estimation, and control purposes across diverse applications.