Parameter Estimation for Two Synthetic Gene Networks: A Case Study

Abstract

In this paper, we use two synthetic gene networks, a transcriptional cascade and a pulse generating network, to study the efficacy of a simple statistical parameter fitting algorithm. The fitting was performed on experimental data and computer-generated data (to test how well the algorithm works under ideal conditions with perfect information). Most of the experimental parameter estimations yielded tight ranges of kinetic values for both gene networks. However, the results using simulated data indicate that the algorithm was able to provide better parameter estimates for the pulse generating network than for the transcriptional cascade. This is likely a result of the larger amount of time-series data available for the pulse generator and its greater level of phenotypical complexity, leading to tighter constraints for optimization. The variation in the magnitudes of the standard deviations between parameter estimates may give an indication of system sensitivity to specific kinetic rate constants. In the future, we also plan to verify the experimental results by constructing network variants and attempting to predict behaviors using values obtained in this study.