High-Performance Agent-Based Modeling Applied to Vocal Fold Inflammation and Repair

Area of Science:

• Computational Biology
• Bioinformatics
• Personalized Medicine

Background:

• Computational biology models accelerate personalized medicine development.
• Estimating treatment success prevents costly, ineffective treatments and side effects.

Purpose of the Study:

• To develop a high-performance Agent-Based Model (ABM) for simulating and visualizing vocal fold inflammation and repair.
• To enable real-time analysis and interaction with complex biological simulations.

Main Methods:

• A novel computational scheme was designed for a 3D Agent-Based Model (ABM) utilizing multi-core CPUs and many-core GPUs.
• Subtasks were parallelized with convolution-based diffusion for enhanced simulation performance.
• A client-server protocol enabled in-situ analysis and visualization during simulation.

Main Results:

• The framework simulated 17 million cells and 1.7 billion data points for human vocal fold injury and repair.
• Simulation, visualization, and remote result delivery were completed in under 7 seconds per iteration (30 real-world minutes).
• The high-performance 3D ABM was verified against empirical vocal fold data, showing accurate biomarker prediction trends.

Conclusions:

• The developed high-performance 3D ABM framework effectively simulates and visualizes complex biological processes in vocal fold repair.
• This tool accelerates the development of personalized medicine by providing fast and accurate treatment success estimations.
• The real-time interactive capabilities offer a powerful approach for understanding and managing vocal fold injuries.