Simulation of a Free Boundary Cell Migration Model through Physics Informed Neural Networks.

Area of Science:

• Biophysics
• Computational Biology
• Cell Biology

Background:

• Computational modeling is crucial for understanding single-cell migration mechanisms.
• Cell migration involves complex interactions of actin polymerization, substrate adhesion, and actomyosin dynamics.

Purpose of the Study:

• To develop a computational model for one-dimensional actomyosin flow in a migrating cell with moving boundaries.
• To apply a deep learning method, specifically a physics-informed neural network, to simulate this biophysical problem.

Main Methods:

• A system of coupled nonlinear partial differential equations was used to model the interplay of biological and physical processes.
• A physics-informed neural network was employed to analyze actin flow, actin concentration, and unknown moving boundaries.
• The model addresses the computational challenges of deformable domains in dynamic simulations.

Main Results:

• Numerical results qualitatively align with existing experimental and computational data.
• The model successfully depicts the intricate interactions governing cell migration.
• The deep learning approach accurately simulates the biophysics of cell migration with moving boundaries.

Conclusions:

• Physics-informed neural networks offer a powerful tool for simulating complex biophysical problems with moving boundaries.
• This model provides a novel computational framework for studying cell migration dynamics.
• The approach demonstrates the potential of AI in advancing our understanding of cellular processes.