Bayesian Computational Modeling: a new tool for understanding the reactive gliósis propagation

RAMOS, ALBERTO; MOFFATT, LUCIANO; AUZMENDI, JERÓNIMO

CABA

Congreso; 2nd FALAN CONGRESS |FALAN-IBROLARC; 2016

FALAN

Ischemic braininjury promotes the acquisition of the reactive astrogliosis phenotype (RA)characterized by an increase of astrocyte volume and secretion ofproinflammatory molecules. Initially focal, RA propagates to distal regions ofthe CNS but the underlying mechanism remains uncertain. Two major possibilitieshave been proposed. The DAMP hypothesis states that the simple diffusion ofDAMPs from the necrotic core is determinant. The Soluble Mediators (SM) hypothesisargues that the subsequent signal carried out by SM secreted by the proximalastrocytes is essential for the RA propagation. We applied the BayesianComputational Modeling (BCM) to distinguish between those mechanisms in thedynamics of the RA propagation. We developed a 1-D mathematical model thatincluded a Markovian Model for the astrocytes activation coupled to a system ofpartial differential equations for the diffusion and binding of DAMPs and SM.Then we found the prior distribution for all the parameters of the system. Weused GFAP-stained rat brain sections at 3 and 7 days post injury by corticaldevascularization and we applied a morphological score to represent theobserved RA gradient. Then we used the model to predict the proportion ofastrocytes in each reactive state. We run a least square non linear fit to findthe most probable combination of parameters and we compared the hypotheses withThermodynamic Integration. We found a strong evidence for the mechanism thatinvolve the secretion of SM.