Sepsis is systemic infection accompanied by an uncontrolled inflammatory response; a condition that can deteriorate rapidly. Early diagnosis is critical for survival. Heart rate variability (HRV), a proposed biomarker for sepsis, predicts its prognosis but is too nonspecific to make a diagnosis. Often HRV is quantified by its power spectra, its variability in the frequency domain; the `high-frequency' component reflects respiratory modulation of vagal nerve activity. Computational deterministic models of the brainstem cardiorespiratory control networks have proposed plausible neural mechanisms for the vago-respiratory coupling. In contrast to HRV, Dynamic Network Analysis (DyNA) and Dynamic Bayesian Network (DyBN) models are highly specific and successful in identifying a `tipping point' in sepsis, i.e. when a controlled inflammatory response becomes uncontrolled but its many variables are hard to measure. Recently, we identified that the brainstem becomes inflamed in endotoxemia. We hypothesize that progressive inflammation is a critical factor in losing HRV, ventilatory pattern variability (VPV), and cardiorespiratory coupling (CRC) associated with sepsis. We propose to build on the strengths of agent-based and computational modeling approaches and perform model-driven experiments to determine how alterations of brainstem neurophysiology in sepsis limit physiologic pattern variability. Our preliminary data show that endotoxemic rats lose CRC progressively in association with proinflammatory cytokines expression first in the nucleus tractus solitarius (nTS) then in the nucleus Ambiguus. Further, consistent with a progressive loss of CRC focal IL-1β microinjections in the nTS uncouples the arterial pulse pressure's influence on respiration leaving RSA intact. The Specific Aims are: 1) to develop DyNa and DyBN models of cytokine expression in brainstem cardiorespiratory control nuclei during septicemia to determine if central and peripheral inflammation patterns, 2) to adapt these models to critically-ill humans at risk for sepsis and probe the robustness of the model by applying therapeutic interventions in rats, and 3) to apply our control model to propose plausible and testable mechanisms for the effects of cytokines on the function of cardiorespiratory control circuitry. Our computational model of the neural control of cardiorespiratory coupling as well as the models defining the interactions among cytokines in tissue inflammation have been applied successfully to other conditions (sympatho-respiratory coupling) or to peripheral tissues (cytokine expression and interaction). Integrating these models will provide cross-scale mechanistic explanations for the loss of RSA and CVC observed during sepsis, identify critical cytokines for therapeutic intervention, and will establish a scientific rationale for using CRC and variability measures as complementary and sensitive biomarkers of sepsis.