Camacho-Gomez Daniel, Sentiero Raffaele, Ventre Maurizio, Garcia-Aznar Jose Manuel
Department of Mechanical Engineering, Multiscale in Mechanical and Biological Engineering (M2BE), Aragon Institute of Engineering Research (I3A), University of Zaragoza, Zaragoza, Spain.
Department of Chemical, Materials and Industrial Production Engineering. University of Naples Federico II, Naples, Italy.
NPJ Syst Biol Appl. 2025 Aug 26;11(1):99. doi: 10.1038/s41540-025-00576-0.
We present a novel computational framework that combines Agent-Based Modeling (ABM) with Reinforcement Learning (RL) using the Double Deep Q-Network (DDQN) algorithm to determine cellular behavior in response to environmental signals. With this approach, the model captures the transduction of environmental cues into biological responses directly from experimental observations, without explicitly predefining cell behavior. This enables the prediction of dynamic, environment-dependent cell behavior and offers a scalable and flexible alternative to traditional rule-based ABM. To illustrate its potential, we present an application to barotactic cell migration data from microfluidic device experiments, where cells adapt their migration behavior based on pressure gradients, demonstrating the model's ability to generalize across varying geometries and pressure configurations. Thus, this approach introduces a novel direction for modeling how cells sense and transduce environmental cues into biological behaviors.
我们提出了一种新颖的计算框架，该框架使用双深度Q网络（DDQN）算法将基于智能体的建模（ABM）与强化学习（RL）相结合，以确定细胞对环境信号的响应行为。通过这种方法，该模型直接从实验观察中捕捉环境线索到生物反应的转导过程，而无需明确预先定义细胞行为。这使得能够预测动态的、依赖环境的细胞行为，并为传统的基于规则的ABM提供了一种可扩展且灵活的替代方法。为了说明其潜力，我们展示了该框架在微流控设备实验中的气压诱导细胞迁移数据上的应用，其中细胞根据压力梯度调整其迁移行为，证明了该模型在不同几何形状和压力配置下的泛化能力。因此，这种方法为模拟细胞如何感知和将环境线索转导为生物行为引入了一个新的方向。
