ParaCells: A GPU Architecture for Cell-Centered Models in Computational Biology.

Area of Science:

• Computational Biology
• Bioinformatics
• Scientific Computing

Background:

• Biological systems exhibit complex hierarchies requiring advanced computational tools.
• Graphics processing units (GPUs) offer parallel computing capabilities suitable for multi-cellular system simulations.
• Scientists often struggle with the mathematical formulation and specialized programming required for GPU simulations.

Purpose of the Study:

• To develop an accessible, cell-centered GPU simulation architecture for modeling multi-cellular biological systems.
• To enable general scientists to perform GPU simulations without deep knowledge of GPU architecture.
• To provide an intuitive interface for complex biological modeling.

Main Methods:

• Introduction of ParaCells, a cell-centered GPU simulation architecture utilizing NVIDIA CUDA.
• Design of ParaCells as a versatile architecture connecting user logic (C++) with NVIDIA CUDA runtime.
• Application of an object-oriented model declaration for broad applicability to biological systems.

Main Results:

• ParaCells achieved significantly faster simulation speeds compared to sequential, OpenMP, and OpenACC implementations.
• Simulation programs developed using ParaCells were cleaner and more readable.
• ParaCells demonstrated versatility in modeling multi-cellular systems through basic biological concept combinations.

Conclusions:

• ParaCells provides an easy-to-use, efficient GPU simulation architecture for computational biology.
• The architecture democratizes GPU computing for scientists, reducing the programming barrier.
• ParaCells facilitates faster, more readable, and versatile multi-cellular system modeling.