Document Type


Submission Date



Introduction: Sepsis is a devastating, costly, and complicated disease. It represents the summation of varied host immune responses in a clinical and physiological diagnosis. Despite extensive research, there is no current mediator-directed therapy, nor a biomarker panel able to categorize disease severity or reliably predict outcome. Although still distant from direct clinical translation, dynamic computational and mathematical models of acute systemic inflammation and sepsis are being developed. Although computationally intensive to run and calibrate, agent-based models (ABMs) are one type of model well suited for this. New analytical methods to efficiently extract knowledge from ABMs are needed. Specifically, machine-learning techniques are a promising option to augment the model development process such that parameterization and calibration are performed intelligently and efficiently.

Methods: We used the Keras framework to train an Artificial Neural Network (ANN) for the purpose of identifying critical biological tipping points at which an in silico patient would heal naturally or require intervention in the Innate Immune Response Agent-Based Model (IIRABM). This ANN, determines simulated patient “survival” from cytokine state based on their overall resilience and the pathogenicity of any active infections experienced by the patient, defined by microbial invasiveness, toxigenesis, and environmental toxicity. These tipping points were gathered from previously generated datasets of simulated sweeps of the 4 IIRABM initializing parameters.

Results: Using mean squared error as our loss function, we report an accuracy of greater than 85% with inclusion of 20% of the training set. This accuracy was independently validated on withheld runs. We note that there is some amount of error that is inherent to this process as the determination of the tipping points is a computation which converges monotonically to the true value as a function of the number of stochastic replicates used to determine the point.

Conclusion: Our method of regression of these critical points represents an alternative to traditional parameter-sweeping or sensitivity analysis techniques. Essentially, the ANN computes the boundaries of the clinically relevant space as a function of the model’s parameterization, eliminating the need for a brute-force exploration of model parameter space. In doing so, we demonstrate the successful development of this ANN which will allows for an efficient exploration of model parameter space.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.