Systematically manipulating T-cell signaling dynamics via multiple model informed open-loop controller design

摘要

A multiple-model approach to open-loop control of T-cell signaling pathways is presented. Mathematical models of the T-cell signaling pathway are used to inform the controller design. The proposed framework employs a model predictive control strategy to reduce the computational complexity of the open loop control problem. Predictions from each model are weighted using adaptive Akaike weights that are iteratively computed for each controller update step based upon the most relevant training data subsets. This process accounts for the fact that models differ in their ability to accurately reflect the system dynamics under different experimental conditions. The algorithm is evaluated in silico and simulations demonstrate how the model weighting strategy more effectively manages the inaccuracies of any single model. Furthermore, the multiple-model control strategy is evaluated in vitro to direct T-cell signaling. The controller-derived input sequence successfully drives the relative concentration of phosphorylated Erk along the desired trajectory when implemented in the laboratory.