Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A novel scalable high performance diffusion solver for multiscale cell simulations

Jose-Luis Estragues-Muñoz, Carlos Alvarez, Arnau Montagud, Daniel Jimenez-Gonzalez, Alfonso Valencia

TL;DR

BioFVM-B delivers an HPC-optimized Finite Volume Method diffusion solver that overcomes memory and serialization bottlenecks of prior solutions (BioFVM and BioFVM-X) by introducing a contiguous microenvironment layout, a 1D MPI domain decomposition with blocking, and vectorized TDMA solvers. The approach achieves up to $196.8\times$ speedups over BioFVM-X and substantial memory reductions, enabling centimeter-scale, organ-level diffusion simulations within multiscale cell models. These results, demonstrated on MareNostrum 5, establish BioFVM-B as a pivotal enabling technology for digital twin disease modeling andPhysiCell-based workflows, with future work targeting accelerators and adaptive blocking heuristics. The work thus significantly advances the practical feasibility of large-scale, atomistic-to-organ diffusion simulations in biomedical HPC.

Abstract

Agent-based cellular models simulate tissue evolution by capturing the behavior of individual cells, their interactions with neighboring cells, and their responses to the surrounding microenvironment. An important challenge in the field is scaling cellular resolution models to real-scale tumor simulations, which is critical for the development of digital twin models of diseases and requires the use of High-Performance Computing (HPC) since every time step involves trillions of operations. We hereby present a scalable HPC solution for the molecular diffusion modeling using an efficient implementation of state-of-the-art Finite Volume Method (FVM) frameworks. The paper systematically evaluates a novel scalable Biological Finite Volume Method (BioFVM) library and presents an extensive performance analysis of the available solutions. Results shows that our HPC proposal reach almost 200x speedup and up to 36% reduction in memory usage over the current state-of-the-art solutions, paving the way to efficiently compute the next generation of biological problems.

A novel scalable high performance diffusion solver for multiscale cell simulations

TL;DR

BioFVM-B delivers an HPC-optimized Finite Volume Method diffusion solver that overcomes memory and serialization bottlenecks of prior solutions (BioFVM and BioFVM-X) by introducing a contiguous microenvironment layout, a 1D MPI domain decomposition with blocking, and vectorized TDMA solvers. The approach achieves up to speedups over BioFVM-X and substantial memory reductions, enabling centimeter-scale, organ-level diffusion simulations within multiscale cell models. These results, demonstrated on MareNostrum 5, establish BioFVM-B as a pivotal enabling technology for digital twin disease modeling andPhysiCell-based workflows, with future work targeting accelerators and adaptive blocking heuristics. The work thus significantly advances the practical feasibility of large-scale, atomistic-to-organ diffusion simulations in biomedical HPC.

Abstract

Agent-based cellular models simulate tissue evolution by capturing the behavior of individual cells, their interactions with neighboring cells, and their responses to the surrounding microenvironment. An important challenge in the field is scaling cellular resolution models to real-scale tumor simulations, which is critical for the development of digital twin models of diseases and requires the use of High-Performance Computing (HPC) since every time step involves trillions of operations. We hereby present a scalable HPC solution for the molecular diffusion modeling using an efficient implementation of state-of-the-art Finite Volume Method (FVM) frameworks. The paper systematically evaluates a novel scalable Biological Finite Volume Method (BioFVM) library and presents an extensive performance analysis of the available solutions. Results shows that our HPC proposal reach almost 200x speedup and up to 36% reduction in memory usage over the current state-of-the-art solutions, paving the way to efficiently compute the next generation of biological problems.
Paper Structure (22 sections, 10 equations, 11 figures, 2 tables, 3 algorithms)

This paper contains 22 sections, 10 equations, 11 figures, 2 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related work
  3. Algorithm complexity and number of operations
  4. BioFVM-B
  5. Distributed Microenvironment class
  6. Diffusion-decay solver for distributed computing
  7. Vectorized solvers
  8. Experimental Setup
  9. Hardware configuration
  10. Software configuration
  11. Benchmarks
  12. Memory and Performance Analysis
  13. Memory Requirement Limitations
  14. Performance results and scalability
  15. Single node scalability
  16. ...and 7 more sections

Figures (11)

  • Figure 1: Comparison of the data structures used by BioFVM and BioFVM-X (top panel) vs. BioFVM-B (bottom panel) for storing substrate values of the microenvironment, under the variable name p_density_vectors within the Microenvironment class. The figure illustrates an example of a cubic microenvironment with dimensions of 2x2x2 voxels, modeling 2 substrates. Positions within the microenvironment are denoted by $(x, y, z)$, while substrate values are indicated as $[id]$.
  • Figure 2: BioFVM-B diffusion-decay solver data distribution. Access patterns are shown for each of the dimension computations. Example of the division of the simulation domain into 4 subdomains that are mapped to different MPI processes. The blocking feature is introduced by BioFVM-B. Illustrated example of 4 MPI processes and factor 1.
  • Figure 3: BioFVM, BioFVM-X and BioFVM-B MareNostrum 5 supercomputer's node usage to allocate different microenvironment sizes.
  • Figure 4: Comparison of time to solution per simulation time step for the diffusion-decay solver across BioFVM, BioFVM-X, and BioFVM-B implementations. Simulations are based on the 4% liver test case. The vertical line indicates the number of physical cores; beyond this point, Hyper-Threading is utilized.
  • Figure 5: BioFVM-B diffusion-decay execution trace in MareNostrum 5 supercomputer analysed with BSC toolsBSC_Tools. Hyper-Threading serialization effects are shown in the threads sharing the same physical core.
  • ...and 6 more figures