Parameter Estimation in Biochemical Models Using Moment Approximations

Biochemical processes exhibiting stochastic fluctuations can be mathematically modeled using the chemical master equation (CME). Because of the natural randomness of the molecular interactions, it is important to infer the model parameters from experimental or synthetic data to ensure that the biochemical process being modeled accurately represents the intended system. Methods for estimating model parameters that require solving the CME are computationally expensive due to its large and potentially infinite size. By contrast, in moment-based approximations, the objective function can be stated as a least squares estimator that avoids solving the CME, thus becoming significantly faster to optimize. We demonstrate the usefulness of this approach by applying it to two case studies from systems biology, also showing that the estimation is accurate with very low relative errors.