Introduction

My name is Salvador de la Torre Gonzalez, and I am a Mathematics and Computer Engineering student at the University of Seville and a Google Summer of Code 2026 contributor working on the project “CARTopiaX: Extending a Next-Generation Platform for Computational Cancer Biology”.

During GSoC 2025, I was part of the Compiler Research team, where I initiated the CARTopiaX framework. This year, the project focuses on extending the platform to reproduce additional biologically relevant phenomena in CAR T-cell therapy research.

Mentors: Luciana Melina Luque, Vassil Vassilev, Lukas Breitwieser

Briefly about CAR-T Cell Therapy and CARTopiaX

Chimeric Antigen Receptor T-cell (CAR-T) therapy is an immunotherapy approach where a patient’s T-cells are engineered to recognize and destroy cancer cells. Although highly successful in blood cancers, CAR-T therapy still faces major challenges in solid tumors due to limited infiltration, immunosuppressive microenvironments, and T-cell exhaustion.

CARTopiaX is an advanced agent-based model designed to study CAR T-cell therapy in solid tumors. Built on BioDynaMo, a high-performance open-source platform for large-scale and modular biological simulations, CARTopiaX enables detailed exploration of complex biological interactions, hypothesis testing and data-driven discovery within tumor microenvironments.

The framework implements the mathematical model proposed in the Nature Scientific Reports paper “In silico study of heterogeneous tumour-derived organoid response to CAR T-cell therapy” and successfully reproduces its experimental trends in silico.

The goal of this GSoC project is to expand CARTopiaX so it can replicate new in vitro experimental observations and biological mechanisms, strengthening its role as a flexible framework for computational cancer biology research.

Implementation Details and Plans

The project will focus on extending CARTopiaX with additional biological mechanisms identified through a literature review of recent CAR T-cell studies.

Possible extension directions

1. Agents

  • Additional immune populations (PBMCs, macrophages, MDSCs): Incorporating alternative immune populations, such as PBMCs (comprising T cells and dendritic cells), macrophages that engulf debris and pathogens, and MDSCs which act as immunosuppressive cells to inhibit T-cell activity.
  • Stromal cells (fibroblasts and CAFs) acting as physical barriers: Including stromal cells like fibroblasts and CAFs (Cancer-Associated Fibroblasts), which contribute to extracellular matrix production and physical immune cell exclusion.
  • Tumor heterogeneity, cancer types, and antigen loss/re-expression: Modeling tumor heterogeneity and variability, which includes differences in antigen expression and the ability of tumor cells to temporarily lose and later regain target antigens to enable immune evasion.

2. Microenvironment

  • Cytokine signaling and immunosuppressive factors: Examining cytokine signaling and immunosuppressive factors, which are signaling proteins that regulate immune activity and can suppress immune responses (e.g., TGF-β, IL-10) to promote tumor evasion.
  • Hypoxia and nutrient deprivation effects: Accounting for hypoxia, which consists of low oxygen regions within tumors caused by limited diffusion and high cellular consumption, along with the associated effects of nutrient deprivation.
  • Vascularization and nutrient diffusion: Simulating vascularization and nutrient diffusion through blood vessels that supply oxygen and nutrients while directly influencing spatial gradients within the tumor microenvironment.
  • ECM viscosity and hydrogel-based barriers: Evaluating ECM viscosity and hydrogel-based barriers, where the density and stiffness of the structural extracellular matrix network hinder immune cell movement.
  • ECM degradation and heat-induced microenvironment modulation: Assessing ECM degradation and heat-induced microenvironment modulation as processes that experimentally modify the extracellular matrix to increase permeability and improve immune cell infiltration.

3. Rules & Dynamics

  • Immune suppression and exhaustion mechanisms: Integrating immune suppression and exhaustion mechanisms, which are processes where immune cells gradually lose their activity due to prolonged stimulation or inhibitory signals.
  • Antigen-dependent recognition and immune evasion: Analyzing antigen-dependent recognition and immune evasion, whereby CAR T-cells recognize tumor cells based on specific antigens while tumors actively evade this by downregulating or losing antigen expression.
  • Chemotaxis and infiltration dynamics: Modeling chemotaxis and infiltration dynamics to capture the directed movement of immune cells toward specific chemical gradients such as cytokines.
  • Macrophage phagocytosis: Incorporating macrophage phagocytosis as the active process by which macrophages engulf and remove dead or damaged cells.
  • ECM permeability evolution: Tracking ECM permeability evolution to understand how continuous changes in tissue structure affect how easily cells can move through the extracellular matrix.
  • Hypoxia-induced necrosis and cell-state transitions: Simulating hypoxia-induced necrosis and cell-state transitions, which involve cell death caused by oxygen deprivation and the subsequent transitions between different cellular states under stress conditions.

The project is organized into four main phases:

Phase 1: Literature Review & Data Acquisition

  • Select a high-impact experimental study from the literature that provides relevant biological and computational insights for model extension.
  • Obtain usable wet-lab datasets associated with the selected study, ensuring they are suitable for computational analysis.
  • Extract quantitative metrics from experimental data to enable model calibration and evaluation.

Phase 2: Model Expansion

  • Add new biological agents, microenvironment components and required to represent the selected CAR T-cell phenomenon.
  • Implement biologically grounded rules and interactions that govern cell behavior and system new dynamics.
  • Adapt and extend the existing CARTopiaX model structure to accurately reproduce the chosen biological setting.

Phase 3: Model Calibration & Optimization

  • Define fitness functions using error metrics (e.g., MSE/RMSE) for model fitting, measuring the discrepancy between simulated outputs and experimental observations.
  • Implement and apply efficient optimization methods for high computational cost scenarios, such as Bayesian optimization and evolutionary algorithms, to explore the parameter space effectively.
  • Reduce computational cost via early stopping, simplified simulations, and staged calibration strategies that progressively refine parameter estimates.
  • Perform parameter estimation to identify the set of model parameters that best reproduce the experimental data.

Phase 4: Validation & Delivery

  • Validate the model using multiple stochastic simulations to ensure robustness and consistency of results.
  • Perform sensitivity analysis and robustness checks to understand parameter influence and model stability.
  • Summarize findings in a structured scientific report oriented towards a future research publication.

Conclusion

This project aims to demonstrate how CARTopiaX can evolve into a flexible and extensible framework capable of reproducing increasingly complex biological phenomena in CAR T-cell therapy research.

By combining real experimental data with large-scale agent-based simulations, the project seeks to support hypothesis-driven exploration in computational oncology while strengthening the connection between in vitro experimentation and predictive in silico modeling.

As both the original author of CARTopiaX and a student working at the intersection of mathematics, computer engineering and computational biology, I am excited to continue expanding the project during GSoC 2026.

  • CARTopiaX
  • Project Description
  • Luque, L.M., Carlevaro, C.M., Rodriguez-Lomba, E. et al. In silico study of heterogeneous tumour-derived organoid response to CAR T-cell therapy. Sci Rep 14, 12307 (2024). DOI
  • Breitwieser, L., Hesam, A., de Montigny, J. et al. BioDynaMo: a modular platform for high-performance agent-based simulation. Bioinformatics 38(2), 453–460 (2022). DOI
  • Breitwieser, L., Hesam, A., Rademakers, F., Gómez Luna, J., and Mutlu, O. High-Performance and Scalable Agent-Based Simulation with BioDynaMo. In Proceedings of the 28th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming (PPoPP ‘23), 174–188 (2023). DOI
  • BioDynaMo Repository
  • My GitHub Profile