Search

Cellular processes are governed by complex networks of interacting genes and proteins. Theoretical molecular biologists attempt to describe these processes via mathematical models by writing biochemical reaction equations. Modellers are building increasingly larger and complex mathematical models to describe these cellular processes, making model evaluation a time consuming and difficult task. The authors describe an automatable process for model evaluation and a software system that implements this process. The software is adaptable to many types of models and is freely available along with all needed data files. The cell cycle control system for budding yeast is known in fine detail and constrained by more than 100 phenotypic observations in mutant strains. As an example, the authors apply their process to a model of cell cycle control in budding yeast containing dozens of regulatory equations and explaining nearly all of the known mutant phenotypes.

DNA synthesis and nuclear division in the developing frog egg are controlled by fluctuations in the activity of M-phase promoting factor (MPF). The biochemical mechanism of MPF regulation is most easily studied in cytoplasmic extracts of frog eggs, for which careful experimental studies of the kinetics of phosphorylation and dephosphorylation of MPF and its regulators have been made. In 1998 Marlovits et al. used these data sets to estimate the kinetic rate constants in a mathematical model of the control system originally proposed by Novak & Tyson. In a recent publication, we showed that a gradient-based optimisation algorithm finds a locally optimal parameter set quite close to the 'Marlovits' estimates. In this paper, we combine global and local optimisation strategies to show that the 'refined Marlovits' parameter set, with one minor but significant modification to the Novak & Tyson equations, is the unique, best-fitting solution to the parameter estimation problem.