(276c) White-Boxing Genetic Circuit Modeling: Absolute TF Binding Free Energies From Fluorescence Measurements
Every genetic circuit model in existence today has been parameterized with proportional data. Fluorescence. Absorbance. Luminescence. These measurements are only proxies for the cell's internal protein concentrations, and yet we use them to record the cell's internal dynamics and to measure the biophysics of its genetic parts. A first-principles biophysical model that uses proportional data will always be left with an unknown proportionality constant, and this unknown factor will depend on many extrinsic and ever-changing variables. Beyond the creation of community-wide measurement standards, we investigated whether a set of properly designed experiments could measure the instrinsic biophysical properties of a genetic part inside the cell, using only proportional fluorescence as the measurement output. We then asked whether this intrinsic knowledge could be used to predict the behavior of a family of synthetic genetic circuits.
Specifically, we develop a high-throughput quantitative approach to measuring the absolute binding free energies of a transcription factor to its DNA operator sites, utilizing only protein reporters and fluorescence measurements. The expression output of the transcription factor's regulated promoter is recorded within a series of rationally designed genetic sandboxes that contain additional non-regulating TF-binding sites. Our first-principles quantitative model shows, and we experimentally demonstrate, that the ratio in fluorescence measurements between two genetic sandboxes yields a direct, absolute relationship to the TF's binding free energy, as measured in kcal/mol.
We validate this approach using three newly engineered repressors, measuring their TF binding free energies, and predicting the behaviors of entire families of genetic circuits. We validate these predictions experimentally and demonstrate that the resulting biophysical model of a TF-regulated promoter can predict its transcription rates across a wide range of TF expression levels. We use these models to identify optimal NOT-gate genetic circuits with maximal dynamic ranges and tunable input thresholds. The affirmative answers to our questions take us two steps closer to creating a genetic circuit engineering discipline that is free of context effects.