Hemaeos Cancer Library

Engineering problem

Hemaeos Cancer Library is a computational tumor-modeling framework for studying heterogeneous cancer-cell populations within spatial biochemical environments. It combines three-dimensional transport, discrete cell agents, simplified invasion mechanisms and reproducible numerical experiments.

The application contains five model families: high-grade serous ovarian carcinoma, breast carcinoma, glioblastoma, colorectal adenocarcinoma and pancreatic ductal adenocarcinoma. They are computational models, not clinically validated digital replicas of those diseases.

Computational architecture

The scientific solver produces archived simulation states containing cell populations, biochemical fields and derived metrics. The browser interface replays those precomputed simulations; it does not execute the complete solver. This separation keeps numerical experiments reproducible while allowing interactive inspection of cells, fields, plots and progression histories.

Mathematical foundations

Reaction-diffusion transport

Biochemical fields follow a spatially heterogeneous transport model,

$$\frac{\partial C}{\partial t}=\nabla\cdot(D(\mathbf{x})\nabla C)-R(C,\mathrm{cells})+S(\mathbf{x}).$$

The terms represent oxygen or nutrient concentration, spatially varying diffusivity, cell-dependent consumption and localized sources or exchange. Cell-centered Cartesian grids use face-based fluxes with Dirichlet or zero-flux Neumann boundaries. Explicit, implicit and quasi-steady field solutions are available, with Picard iteration for nonlinear consumption and checks for positivity and conservation.

Agent-based cell dynamics

Each tumor-cell agent stores position, clone identity, simulated biomarker indices and cell state. Rules cover proliferation, cell-cycle progression, quiescence, hypoxia-induced necrosis, nutrient-stress apoptosis, necrotic clearance, contact inhibition and simplified displacement. Cells consume local field quantities while those fields influence cell state and behavior.

Biomarker values are simulated dimensionless indices, not measured clinical expression. Clone-dependent parameters can vary growth, consumption, receptor distributions and invasive behavior.

Cancer model families

High-grade serous ovarian carcinoma

The ovarian model includes oxygen-limited spheroid growth, hypoxia, necrotic-core development, nutrient-stress apoptosis, necrotic clearance and heterogeneous FOLR1, MSLN, MUC16 and NECTIN4 indices across seeded-lesion scenarios.

Breast carcinoma

The breast model uses tumor-interface geometry, clone-dependent invasion, ERBB2-related simulated phenotypes, ESR1, PGR and ERBB2 indices and crowding-driven local displacement.

Glioblastoma

The glioblastoma model includes region-dependent diffusion, a high-diffusivity corridor, a simplified vessel-exchange source and EGFR-index-associated invasive behavior.

Colorectal adenocarcinoma

The colorectal model represents mucosa and wall transport regions, region-dependent nutrient diffusivity, CTNNB1-index-associated invasion and interface-crossing measurements.

Pancreatic ductal adenocarcinoma

The pancreatic model investigates stromal transport barriers, reduced regional diffusivity, oxygen limitation, necrosis and local or invasive progression.

Invasion and progression metrics

Geometric tissue interfaces support invasion-depth and boundary-crossing measurements for local and invasive scenarios. These are numerical geometric metrics and are not equivalents of clinical FIGO or TNM staging.

Engineering case studies

• Ovarian receptor heterogeneity: varies receptor-index distribution width and records changes in simulated heterogeneity.
• Pancreatic stromal barrier: compares diffusivity factors of 1.00, 0.15 and 0.05, with 0, 96 and 168 recorded necrosis entries under the assumed parameters.
• Glioblastoma localized exchange: varies a simplified local vessel-exchange coefficient and observes the oxygen field response.
• Breast invasion speed: compares 0, 8 and 40 µm/h, producing invasion depths of 0, 13.9 and 47.4 µm and crossing counts of 0, 31 and 86 in the corresponding numerical runs.
• Colorectal regional transport: examines how region-specific nutrient diffusivity changes simulated tumor-cell exposure.

These are numerical outcomes under assumed parameters, not experimental biological findings.

Numerical verification

The framework uses deterministic seeds, population accounting, conservation and positivity checks, time-step and grid-spacing sensitivity studies and archive-integrity checks. These tests verify implementation behavior but do not establish biological validation or general numerical convergence beyond the tested configurations.

Scientific interface

The live application provides 3D tumor-cell visualization, state and clone coloring, biomarker overlays, oxygen, nutrient and pH-proxy slices, cell and voxel inspection, archived playback, population plots, invasion metrics and case-study comparisons. Its source repository is private.

Engineering applications

The project supports computational oncology experiments involving tumor microenvironments, transport barriers, heterogeneous agent populations, numerical study design and scientific visualization.