SciEPublish

Investigating the Impact of Measurement Variance on Parameterization: A Case Study of Synthetic Gene Circuits

Article Open Access

Investigating the Impact of Measurement Variance on Parameterization: A Case Study of Synthetic Gene Circuits

Author Information
1
Cain Department of Chemical Engineering, Louisiana State University, Baton Rouge, LA 70803, USA
2
Baton Rouge Magnet High School, Baton Rouge, LA 70806, USA
3
Celia Scott Weatherhead School of Public Health & Tropical Medicine, Tulane University, New Orleans, LA 70112, USA
*
Authors to whom correspondence should be addressed.

Received: 21 September 2026 Revised: 23 October 2026 Accepted: 02 November 2026 Published: 16 November 2026

Creative Commons

© 2026 The authors. This is an open access article under the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/).

Views:34
Downloads:25
Synth. Biol. Eng. 2026, 4(4), 10017; DOI: 10.70322/sbe.2026.10017
ABSTRACT: Ordinary differential equation (ODE) models are widely used to characterize synthetic gene circuits, but estimating model parameters from fluorescence-based gene expression data can be challenging. This study examines whether measurement variance should be incorporated into objective functions and whether objective function formulation can benefit parameter identifiability. We fitted mechanistic models of a toehold-switch AND gate and a CRISPR interference (CRISPRi) regulator to cell-free experimental measurements using unweighted least squares, inverse-variance weighting, and relative-variance weighting under raw and noise-augmented data. The three objectives produced similar trajectory fits, with relative-variance weighting showing modest robustness under higher noise, while inverse-variance weighting was less consistent. Profile-likelihood analysis showed that residual weighting did not resolve practical non-identifiability. These findings were then cross validated using an independent STAR circuit. Consistently, objective function choice had limited impact on fitting or parameter identifiability. To further examine parameter identifiability, we reduced the original eight-parameter STAR model to a three-state, four-parameter model and found that reduced model complexity can lead to local parameter identifiability, at the cost of reduced approximation accuracy to the time-course measurements. While this study focused on three representative gene circuits, we anticipate future studies across a broader range of circuit architectures to establish more general conclusions regarding the impact of measurement variance on model parameterization.
Keywords: Gene circuits; Model parameterization; Measurement variance; Profile likelihood; Structural identifiability; Synthetic biology; Cell-free systems
TOP