STEPS 4.0

STEPS 4.0 performs fast, memory-efficient stochastic spatial reaction-diffusion simulations and coupled deterministic membrane potential calculations to model nanoscale molecular dynamics in neuronal and whole-cell systems.

Key Features:

• Parallel Stochastic Reaction-Diffusion Solver: Distributes computation to handle complex stochastic, spatial reaction-diffusion processes across large-scale systems.
• Deterministic Membrane Potential Solver: Couples a deterministic membrane potential solver with the stochastic solver to simulate membrane dynamics relevant to neuronal activity such as calcium burst activity in Purkinje neurons.
• Distributed Mesh Solution: Implements a novel distributed mesh architecture that optimizes data layout and algorithm design to reduce memory footprint and enable distributed computation.
• Enhanced Memory Efficiency: Achieves a reduction in per-core memory consumption by more than a factor of 30 while maintaining or improving performance and scalability.
• Scalability and HPC Parallelism: Facilitates massively parallel simulations on modern high-performance computing (HPC) clusters.
• Benchmarking: Has been benchmarked against three published models with varying complexities, demonstrating capability on both simple spatial stochastic models and more intricate coupled systems.
• Software Development Practices: Updates to the codebase and development environment follow modern software engineering principles to support future development.

Scientific Applications:

• Neuronal system modeling: Simulates molecular dynamics and membrane potential interactions in neuronal systems, including calcium burst activity in Purkinje neurons.
• Whole-cell spatial modeling: Enables stochastic, spatial reaction-diffusion simulations that preserve nanoscale detail within macroscopic whole-cell contexts.
• Large-scale simulations on HPC: Supports memory-efficient, massively parallel simulation of complex biological models on HPC clusters.
• Model benchmarking and validation: Applied to published models of varying complexity to assess performance, scalability, and memory usage.

Methodology:

Combines a parallel stochastic reaction-diffusion solver, a coupled deterministic membrane potential solver, and a novel distributed mesh with optimized data layout and algorithm designs, and has been benchmarked against three published models.