Data-driven modeling and prediction of microglial cell dynamics in the ischemic penumbra

Sara Amato,Andrea Arnold

2024-10-20

Abstract:Neuroinflammation immediately follows the onset of ischemic stroke. During this process, microglial cells are activated in and recruited to the tissue surrounding the irreversibly injured infarct core, referred to as the penumbra. Microglial cells can be activated into two distinct phenotypes; however, the dynamics between the detrimental M1 phenotype and beneficial M2 phenotype are not fully understood. Using phenotype-specific cell count data obtained from experimental studies on middle cerebral artery occlusion-induced stroke in mice, we employ sparsity-promoting system identification techniques combined with Bayesian statistical methods for uncertainty quantification to generate continuous and discrete-time predictive models of the M1 and M2 microglial cell dynamics. The resulting data-driven models include constant and linear terms but do not include nonlinear interactions between the cells. Results emphasize an initial M2 dominance followed by a takeover of M1 cells, capture potential long-term dynamics of microglial cells, and suggest a persistent inflammatory response.

Cell Behavior,Quantitative Methods,Applications,Computation

What problem does this paper attempt to address?

The paper attempts to address the issue of the dynamic behavior and transformation mechanisms of microglia in the penumbra region following ischemic stroke. Specifically, the paper focuses on the dynamic changes between the M1 (harmful) and M2 (beneficial) phenotypes of microglia and how these changes affect the neuroinflammatory process. Through experimental data and computational modeling techniques, the authors aim to generate data-driven predictive models to better understand microglial behavior and provide a theoretical basis for clinical interventions.

### Background of the Paper
- **Neuroinflammation**: Neuroinflammation occurs immediately after ischemic stroke, with microglia being activated and recruited to the tissue surrounding the irreversibly damaged core area (referred to as the infarct core), known as the penumbra.
- **Microglial Phenotypes**: Microglia can be activated into two different phenotypes—M1 and M2. The M1 phenotype is generally associated with harmful effects, while the M2 phenotype is associated with beneficial effects. The dynamic relationship between these two phenotypes is not yet fully understood.
- **Research Objective**: To generate data-driven predictive models through experimental data and computational modeling techniques to help understand microglial behavior after ischemic stroke and infer their long-term dynamics.

### Research Methods
- **Data Source**: Data on M1 and M2 microglial counts obtained from experimental studies, primarily involving the middle cerebral artery occlusion (MCAO) induced stroke model in mice.
- **Modeling Methods**:
- **SINDy+DE Method**: Using the Sparse Identification of Nonlinear Dynamics (SINDy) algorithm combined with Bayesian statistical methods for uncertainty quantification to generate continuous-time dynamic models.
- **ESINDy+DBN Method**: Using the Ensemble SINDy (ESINDy) algorithm to identify the most likely Dynamic Bayesian Network (DBN) structure to generate discrete-time dynamic models.

### Main Contributions
- **New Data-Driven Models**: Proposed two new data-driven models, one being a continuous-time dynamic system model and the other a discrete-time model.
- **Numerical Framework**: Developed a numerical framework combining SINDy model identification and Bayesian statistical techniques to generate robust data-driven models suitable for prediction and uncertainty quantification.
- **Biological Understanding**: Utilized the generated predictive models to help understand the dynamic behavior of microglia after MCAO-induced stroke and infer their long-term dynamics.

### Results
- **SINDy+DE Method**: Generated a continuous-time dynamic model containing only constant and linear terms, without nonlinear interaction terms. The model results indicate that the M2 phenotype dominates initially, followed by a gradual dominance of the M1 phenotype.
- **ESINDy+DBN Method**: Generated a discrete-time dynamic model, also containing only constant and linear terms. The model results support the initial dominance of the M2 phenotype and the later dominance of the M1 phenotype.

### Conclusion
- **Short-term and Long-term Dynamics**: The model results emphasize the initial dominance of the M2 phenotype and the later dominance of the M1 phenotype, capturing the potential long-term dynamics of microglia and suggesting a sustained inflammatory response.
- **Clinical Significance**: These models help understand microglial behavior after ischemic stroke, providing a theoretical basis for clinical interventions, particularly for treatment strategies targeting the penumbra region.