Sparse-Grid-Based Adaptive Model Predictive Control of HL60 Cellular Differentiation

Noble, Sarah L.; Wendel, L. E.; Donahue, Maia M.; Buzzard, Gregery T.; Rundell, Ann E.

doi:10.1109/tbme.2011.2174361

Cited by 4 publications

(10 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A few notable works that used feedback control approaches paired in silico control with experimental corroboration include controlling cell differentiation via nonlinear MPC [13], [14] and long-term feedback control of gene expression in yeast [15], [16]. This study augments this literature base by providing a novel and experimentally-corroborated methodology to enable the successful design and implementation of open-loop control strategies.…”

Section: Discussionmentioning

confidence: 91%

“…Sparse grids have been used as computational cost-cutting tools for control applications in systems biology [14], [30] by serving as surrogates for slow-evaluating models and objective functions to allow rapid screening of the design space.…”

Section: Resultsmentioning

confidence: 99%

“…in silico feedback) control to direct HL60 cell differentiation through periodic boluses of a differentiation-inducing agent determined by nonlinear model predictive control (MPC). In 2012, they revised the initial approach to improve the transient response of the differentiating cells over 20 days by using a multi-scenario adaptive model predictive control [14]. Uhlendorf et al [15] applied open-loop (i.e.…”

Section: Introductionmentioning

confidence: 99%

“…Multiple models, or scenarios, have been previously used to improve robustness to parametric uncertainty in closed-loop model-based control [14], [22]–[25]. The approach proposed by Rao et al [24] computed control inputs by weighting multiple step-response models using a Bayesian algorithm to control hemodynamic variables in hypertensive subjects.…”

Section: Introductionmentioning

confidence: 99%

“…The recursive weighting system was effective at eliminating the “hard switch” between controllers. Noble et al [14] employed adaptive nonlinear MPC based on multiple data-consistent parameter characterizations to manipulate cell differentiation experimentally. Control inputs were chosen such that the average tracking error and resource efficiency were optimized, which resulted in superior controller performance over single-model MPC.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Multiple Model-Informed Open-Loop Control of Uncertain Intracellular Signaling Dynamics

et al. 2014

View full text Add to dashboard Cite

Computational approaches to tune the activation of intracellular signal transduction pathways both predictably and selectively will enable researchers to explore and interrogate cell biology with unprecedented precision. Techniques to control complex nonlinear systems typically involve the application of control theory to a descriptive mathematical model. For cellular processes, however, measurement assays tend to be too time consuming for real-time feedback control and models offer rough approximations of the biological reality, thus limiting their utility when considered in isolation. We overcome these problems by combining nonlinear model predictive control with a novel adaptive weighting algorithm that blends predictions from multiple models to derive a compromise open-loop control sequence. The proposed strategy uses weight maps to inform the controller of the tendency for models to differ in their ability to accurately reproduce the system dynamics under different experimental perturbations (i.e. control inputs). These maps, which characterize the changing model likelihoods over the admissible control input space, are constructed using preexisting experimental data and used to produce a model-based open-loop control framework. In effect, the proposed method designs a sequence of control inputs that force the signaling dynamics along a predefined temporal response without measurement feedback while mitigating the effects of model uncertainty. We demonstrate this technique on the well-known Erk/MAPK signaling pathway in T cells. In silico assessment demonstrates that this approach successfully reduces target tracking error by 52% or better when compared with single model-based controllers and non-adaptive multiple model-based controllers. In vitro implementation of the proposed approach in Jurkat cells confirms a 63% reduction in tracking error when compared with the best of the single-model controllers. This study provides an experimentally-corroborated control methodology that utilizes the knowledge encoded within multiple mathematical models of intracellular signaling to design control inputs that effectively direct cell behavior in open-loop.

show abstract