Publications

4D hybrid model interrogates agent-level rules and parameters driving hiPS cell colony dynamics

JS Yu, B Lyons, SM Rafelski, JA Theriot, N Bagheri, GT Johnson. bioRxiv. DOI: 10.1101/2024.08.12.607546

Iterating between data-driven research and generative computational models is a powerful approach for emulating biological systems, testing hypotheses, and gaining a deeper understanding of these systems. We developed a hybrid agent-based model (ABM) that integrates a Cellular Potts Model (CPM) designed to investigate cell shape and colony dynamics in human induced pluripotent stem cell (hiPS cell) colonies. This model aimed to first mimic and then explore the dynamics observed in real-world hiPS cell cultures. Initial outputs showed great potential, seeming to mimic small colony behaviors relatively well. However, longer simulations and quantitative comparisons revealed limitations, particularly with the CPM component, which lacked long-range interactions that might be necessary for accurate simulations. This challenge led us to thoroughly examine the hybrid model’s potential and limitations, providing insights and recommendations for systems where cell-wide mechanics play significant roles. The CPM supports 2D and 3D cell shapes using a Monte Carlo algorithm to prevent cell fragmentation. Basic “out of the box” CPM Hamiltonian terms of volume and adhesion were insufficient to match live cell imaging of hiPS cell cultures. Adding substrate adhesion resulted in flatter colonies, highlighting the need to consider environmental context in modeling. High-throughput parameter sweeps identified regimes that produced consistent simulated shapes and demonstrated the impact of specific model decisions on emergent dynamics. Full-scale simulations showed that while certain agent rules could form a hiPS cell monolayer in 3D, they could not maintain it over time. Our study underscores that “out of the box” 3D CPMs, which do not natively incorporate long-range cell mechanics like elasticity, may be insufficient for accurately simulating hiPS cell and colony dynamics. To address this limitation, future work could add mechanical constraints to the CPM Hamiltonian or integrate global agent rules. Alternatively, replacing the CPM with a methodology that directly represents cell mechanics might be necessary. Documenting and sharing our model development process fosters open team science and supports the broader research community in developing computational models of complex biological systems.

Incorporating temporal information during feature engineering bolsters emulation of spatio-temporal emergence

JY Cain, JI Evarts, JS Yu, N Bagheri. Bioinformatics. DOI: 10.1093/bioinformatics/btae131

Emergent biological dynamics derive from the evolution of lower-level spatial and temporal processes. A long-standing challenge for scientists and engineers is identifying simple low-level rules that give rise to complex higher-level dynamics. High-resolution biological data acquisition enables this identification and has evolved at a rapid pace for both experimental and computational approaches. Simultaneously harnessing the resolution and managing the expense of emerging technologies—e.g. live cell imaging, scRNAseq, agent-based models—requires a deeper understanding of how spatial and temporal axes impact biological systems. Effective emulation is a promising solution to manage the expense of increasingly complex high-resolution computational models. In this research, we focus on the emulation of a tumor microenvironment agent-based model to examine the relationship between spatial and temporal environment features, and emergent tumor properties. Despite significant feature engineering, we find limited predictive capacity of tumor properties from initial system representations. However, incorporating temporal information derived from intermediate simulation states dramatically improves the predictive performance of machine learning models. We train a deep-learning emulator on intermediate simulation states and observe promising enhancements over emulators trained solely on initial conditions. Our results underscore the importance of incorporating temporal information in the evaluation of spatio-temporal emergent behavior. Nevertheless, the emulators exhibit inconsistent performance, suggesting that the underlying model characterizes unique cell populations dynamics that are not easily replaced.

Model design choices impact biological insight: Unpacking the broad landscape of spatial-temporal model development decisions

JS Yu, N Bagheri. PLOS Computational Biology. DOI: 10.1371/journal.pcbi.1011917

Computational models enable scientists to understand observed dynamics, uncover rules underlying behaviors, predict experimental outcomes, and generate new hypotheses. There are countless modeling approaches that can be used to characterize biological systems, further multiplied when accounting for the variety of model design choices. Many studies focus on the impact of model parameters on model output and performance; fewer studies investigate the impact of model design choices on biological insight. Here we demonstrate why model design choices should be deliberate and intentional in context of the specific research system and question. In this study, we analyze agnostic and broadly applicable modeling choices at three levels—system, cell, and environment—within the same agent-based modeling framework to interrogate their impact on temporal, spatial, and single-cell emergent dynamics. We identify key considerations when making these modeling choices, including the (i) differences between qualitative vs. quantitative results driven by choices in system representation, (ii) impact of cell-to-cell variability choices on cell-level and temporal trends, and (iii) relationship between emergent outcomes and choices of nutrient dynamics in the environment. This generalizable investigation can help guide the choices made when developing biological models that aim to characterize spatial-temporal dynamics.

Mapping CAR T-cell design space using agent-based models

AN Prybutok, JS Yu, JN Leonard, N Bagheri. Frontiers in Molecular Biosciences. DOI: 10.3389/fmolb.2022.849363

Chimeric antigen receptor (CAR) T-cell therapy shows promise for treating liquid cancers and increasingly for solid tumors as well. While potential design strategies exist to address translational challenges, including the lack of unique tumor antigens and the presence of an immunosuppressive tumor microenvironment, testing all possible design choices in vitro and in vivo is prohibitively expensive, time consuming, and laborious. To address this gap, we extended the modeling framework ARCADE (Agent-based Representation of Cells And Dynamic Environments) to include CAR T-cell agents (CAR T-cell ARCADE, or CARCADE). We conducted in silico experiments to investigate how clinically relevant design choices and inherent tumor features—CAR T-cell dose, CD4+:CD8+ CAR T-cell ratio, CAR-antigen affinity, cancer and healthy cell antigen expression—individually and collectively impact treatment outcomes. Our analysis revealed that tuning CAR affinity modulates IL-2 production by balancing CAR T-cell proliferation and effector function. It also identified a novel multi-feature tuned treatment strategy for balancing selectivity and efficacy and provided insights into how spatial effects can impact relative treatment performance in different contexts. CARCADE facilitates deeper biological understanding of treatment design and could ultimately enable identification of promising treatment strategies to accelerate solid tumor CAR T-cell design-build-test cycles.

Modular microenvironment components reproduce vascular dynamics de novo in a multi-scale agent-based model

JS Yu, N Bagheri. Cell Systems. DOI: 10.1016/j.cels.2021.05.007

Cells do not exist in isolation; they continuously act within and react to their environment. And this environment is not static; it continuously adapts and responds to cells. Here, we investigate how vascular structure and function impact emergent cell population behavior using an agent-based model (ABM). Our ABM enables researchers to “mix and match” cell agents, subcellular modules, and microenvironment components ranging from simple nutrient sources to complex, realistic vascular architectures that accurately capture hemodynamics. We use this ABM to highlight the bilateral relationship between cells and nearby vasculature, demonstrate the effect of vascular structure on environmental heterogeneity, and emphasize the non-linear, non-intuitive relationship between vascular function and the behavior of cell populations over time. Our ABM is well suited to characterizing in vitro and in vivo studies, with applications from basic science to translational synthetic biology and medicine.

Agent-based models predict emergent behavior of heterogeneous cell populations in dynamic microenvironments

JS Yu, N Bagheri. Frontiers in Bioengineering and Biotechnology. DOI: 10.3389/fbioe.2020.00249

Computational models are most impactful when they explain and characterize biological phenomena that are non-intuitive, unexpected, or difficult to study experimentally. Countless equation-based models have been built for these purposes, but we have yet to realize the extent to which rules-based models offer an intuitive framework that encourages computational and experimental collaboration. We develop ARCADE, a multi-scale agent-based model to interrogate emergent behavior of heterogeneous cell agents within dynamic microenvironments and demonstrate how complexity of intracellular metabolism and signaling modules impacts emergent dynamics. We perform in silico case studies on context, competition, and heterogeneity to demonstrate the utility of our model for gaining computational and experimental insight. Notably, there exist (i) differences in emergent behavior between colony and tissue contexts, (ii) linear, non-linear, and multimodal consequences of parameter variation on competition in simulated co-cultures, and (iii) variable impact of cell and population heterogeneity on emergent outcomes. Our extensible framework is easily modified to explore numerous biological systems, from tumor microenvironments to microbiomes.