Recursive Bayesian estimation of stochastic rate constants from heterogeneous cell populations

Robust estimation of kinetic parameters of intracellular processes requires large amounts of quantitative data. Due to the high uncertainty of such processes and the fact that recent single-cell measurement techniques have limited resolution and dimensionality, estimation should pool recordings of multiple cells of an isogenic cell population. However, experimental results have shown that several factors such as cell volume or cell-cycle stage can drastically affect signaling as well as protein expression, leading to inherent heterogeneities in the cell population measurements. Here we present a recursive Bayesian estimation procedure for stochastic kinetic model calibration using heterogeneous cell population data. While obtaining optimal estimates for the rate constants, this approach allows to reconstruct missing species as well as to quantitatively capture extrinsic variability. The proposed algorithm is applied to a model of the osmo-stress induced MAPK Hog1 activation in the cytoplasm and its translocation to the nucleus.