February 23, 2026
- Demonstrates scalable, GPU-accelerated adhesive dynamics across anatomically realistic vasculature
- Tracks thousands of circulating tumor cells concurrently with strong reproducibility guarantees
- Achieves tissue-scale resolution at a fraction of the memory cost of fully explicit simulations
- Showcases multi-window scalability on a U.S. exascale supercomputer
We are excited to announce the publication of a new article from the Randles Lab, “High-throughput adaptive physics refinement for tissue-scale adhesive dynamics,” authored by Aristotle Martin, William Ladd, Runxin Wu, and Amanda Randles. The paper is published in Journal of Computational Science and is now available online.
This work introduces a high-throughput extension of Adaptive Physics Refinement for Adhesive Dynamics (APR-AD), enabling tissue-scale simulation of receptor-mediated cell adhesion at submicrometer resolution. By combining heterogeneous CPU–GPU execution, a multi-window wall-resolving formulation, and a one-way coupling strategy between bulk flow and fine-scale adhesion domains, the new framework makes it possible to track thousands of circulating tumor cells concurrently without sacrificing physical fidelity.
The authors demonstrate deterministic, decomposition-invariant reproducibility of GPU-accelerated adhesive dynamics and introduce a communication-free APR mode for steady-state flows that transforms the fine-scale window phase into an embarrassingly parallel workload. Scaled on the Aurora supercomputer, the approach tracks over 3,000 circulating tumor cells simultaneously while reducing memory requirements by approximately 15× compared to fully explicit simulations.
Together, these advances expand APR-AD from a single-cell feasibility method into a high-throughput computational microscope for studying cancer transport and other receptor-mediated biological processes across realistic vascular geometries.