NeuroGPU: Accelerating multi-compartment, biophysically detailed neuron simulations on GPUs.

Area of Science:

• Computational neuroscience
• Biophysics
• High-performance computing

Background:

• Neuronal membrane potential relies on complex, multi-scale biophysical processes.
• Accurate neuron models require detailed simulations but face parameter fitting challenges.
• Current central processing unit (CPU)-based simulations are time-consuming, limiting model quality.

Purpose of the Study:

• Introduce NeuroGPU, a novel simulation environment utilizing graphics processing units (GPUs).
• Accelerate the simulation of biophysically detailed neuron models.
• Facilitate efficient parameter exploration and model optimization.

Main Methods:

• Leverages the parallel processing capabilities of GPUs for neuronal simulations.
• Designed for high-throughput parameter tuning by running multiple model instances concurrently.
• Optimized for multi-GPU utilization to maximize computational speed-up.

Main Results:

• Achieves 10-200x speed-up over single-core NEURON simulations.
• Demonstrates up to 800x speed-up with multi-GPU configurations for parameter exploration.
• Enables rapid, large-scale parameter space exploration and accurate model fitting.

Conclusions:

• NeuroGPU offers the fastest platform for simulating detailed multi-compartment neuron models.
• Accessible on common computing systems, empowering a wider scientific community.
• Facilitates advanced research in computational neuroscience through accelerated simulations.