Agent-based modeling of tumor microenvironments

Built PhysiCell scripting infrastructure on the Zaratan HPC cluster and replicated Heiland et al. (18 runs, 8 of 9 checkpoints within tolerance). Designed multi-cell-type computational domains (cancer, cancer-immune, and immune-only clusters) and wrote Python scripts to compute cell-type densities and distance-based metrics from simulation output. Implemented trackable cell agents to quantify distance to defined regions, regional residence times, and local cell density over time, ran parameter sweeps of 100 to 1,000 parallel simulations varying cell motility speed and adhesion strength, aggregated results across runs for model calibration, and implemented a spatial niche calculation pipeline.

The question

How can multi-cellular spatial dynamics in tumor microenvironments be simulated reproducibly at scale on high-performance computing infrastructure?

What I did

Built PhysiCell scripting infrastructure for batch simulation workflows on the University of Maryland Zaratan HPC cluster. Replicated the published Heiland et al. agent-based model across 18 simulation runs, verifying 8 of 9 quantitative checkpoints within tolerance, and implemented a spatial niche calculation pipeline for downstream spatial analysis.

Methods and tools

PhysiCell (C++), Python analysis pipelines, Slurm job orchestration on the Zaratan HPC cluster, and spatial neighborhood statistics.