Designing a hybrid multiscale computational model to explore multicell spatial patterning driven by signaling-induced differential adhesion; Redesigning the modeling process: measures that justify the epistemic value of rule-based algorithms

Computational rule-based algorithms map how behaviors of individuals in a system collectively drive population-level outcomes. The technical project applies this form of modeling to mechanistically explore how molecular events inside cells drive cell-cell interactions that collectively have tissue- and organ-level impacts on disease. The predictions generated by this rule-based model fundamentally depend on modeler-defined rules that link cause and effect within the system. Due to the high dependence of model conclusions on modeler assumptions, the science, technology, and society (STS) research explores the extent to which these rule-based algorithms can lend epistemic value in biomedical research. Moreover, the STS research identifies modeling practices that can mitigate unintended consequences associated with misinterpretation of model results and assumptions. We rigorously integrated these practices in the development of the technical model to effectively explore a critical biological process.
While entirely different in outcome, the early stages of development and the progression of several severe cancers hinge upon the same underlying process: the ability of individual cells to collectively form multicell patterns. Understanding how molecular mechanisms inside cells and interactions between cells drive multicell patterning will inform treatable targets for various developmental and disease pathophysiologies. The technical project created and validated a computational model to mechanistically explore how differential adhesion drives multicell patterning. The model was intrinsically “rule-based,” meaning we defined a ruleset that defines how individual cells act based on their current state and surrounding environment. Based on individual cells executing these rules over time, the model simulated patterns forming at the multicell-level.
Systematic exploration of the multiscale computational model identified critical parameters that drive multicell patterning and how specific combinations of these parameters generate distinct patterns. We found that the adhesion strength between cells and the probability of cell-cell signaling inducing adhesion behaviors strongly influence pattern formation. Covarying these parameters in cell types indicated how specific combinations of these parameters gave rise to core/pole, core/shell, striped, bull’s eye, and soccer ball patterns. These results describe targetable drivers of pattern formation and demonstrate how we can apply this model to engineer multicell aggregates with a desired final pattern.
However, the conclusions of this rule-based model heavily rely on intrinsic assumptions, calling into question whether the model pinpoints the actual mechanism for a specific pattern to form or simply one possible way in which this phenomenon could arise. Moreover, the rules for cell movement and cell clustering are highly generalizable to interactions between agents in non-cellular contexts. Due to the high dependence on intrinsic biases and generalizability of rule-based algorithms, these models can have unintended consequences in several domains. The general Social Construction of Technology (SCOT) framework from Bijker, Hughes, and Pinch maps the negotiation between engineers and several social groups in the application of proposed technologies. The STS research analyzed case studies of rule-based model development with a SCOT framework to identify modeling practices that clarify the negotiation process between the engineer and user groups when developing and deploying rule-based algorithms.
Based on analysis of several case studies, the STS thesis concludes that rule-based algorithms can only yield explanations of how possibly an emergent phenomenon arises. Accordingly, the STS research expands on how modelers can still lend epistemic value to the scientific community by framing conclusions within a “how-possibly” context. Specifically, we propose that modelers conduct robustness analysis during model development to thoroughly disclose how model conclusions hinge upon model assumptions. Moreover, we find that diverse and empathetic modeling groups are more likely to create models that minimize bias.
Conclusions from the STS work were tightly coupled with the technical research. Robustness analysis and collaboration with a diverse lab group helped thoroughly explore the biological mechanisms that drive tissue-level patterning. This STS work will guide the effective dissemination of several computational models in systems biology research.

School of Engineering and Applied Science
Bachelor of Science in Biomedical Engineering
Technical Advisor: Shayn Peirce-Cottler
STS Advisor: Catherine Baritaud