A screening method for dimensionality reduction in biochemical reaction system calibration

Estimating the rate constants of a biochemical reaction model of cellular function is an important, albeit computationally intensive, problem in systems biology. In this paper, a variance-based sensitivity analysis approach is proposed, which can be used, as a pre-screening step, to identify parameters in a biochemical reaction system that do not appreciably influence the cost of estimation and, therefore, whose values cannot be precisely determined by parameter estimation. By only estimating the remaining parameters, appreciable qualitative and quantitative improvements can be achieved. A subset of a well-known biochemical reaction model of the EGF/ERK signaling pathway is used to illustrate the benefits achieved by the proposed method.