Towards optimal virtual patients: an online adaptive control approach.

The application of biosimulation to drug discovery and optimization is enhanced by applying in silico disease models that capture reported heterogeneity in patient clinical phenotypes. Using such a diverse cohort of virtual patients improves the robustness of the in silico analysis and allows critical hypothesis testing to explore key knowledge gaps. The rapid development of a diverse virtual patient cohort exhibiting appropriate steady-state and dynamic behaviors subject to a wide spectrum of stimuli is challenging due to the complexity of the mathematical representation of the biological system, rendering manual parameter tuning infeasible. In this paper, we present an online adaptive control technique, based on model reference adaptive control (MRAC), to optimally auto-tune model parameters for a virtual patient population in order to meet the desired stimulus-response constraints. We validate the efficacy of the control scheme on the Entelos Metabolism PhysioLab platform by automatically generating a cohort of validated virtual patients suitable for in silico research.