Hormone biosensor optimization with a novel biophysical modeling pipeline

me biosensors is critical for advancing non-invasive point-of-care diagnostics. Allosteric transcription factors (aTFs) innately regulate gene expression in microbes by binding DNA in response to the presence or absence of small molecule ligands. Hormones such as progesterone (PRG) and estrogen are key reproductive health biomarkers and present ideal targets for aTF-based sensors. The Galagan lab recently developed the first PRG-responsive biosensor using SRTF1, an aTF from P. simplex, whose DNA-binding affinity is modulated by PRG. This sensor operates via Förster Resonance Energy Transfer (FRET), enabling quantitative detection of PRG concentration. BoltzNet, a biophysical neural network developed in our lab, predicts aTF-DNA binding affinities in E. coli based on DNA sequence, providing a powerful tool for biosensor design. However, no framework currently links these computational predictions to experimental sensor performance. This research validates BoltzNet for non-E. coli TFs and establishes a bi-directional modeling pipeline to predict and optimize biosensor function. By expanding existing predictive modeling frameworks and conducting biosensor characterization and experimental validation, this research enables rational in silico design of TF-DNA biosensors. Ultimately, this project seeks to demonstrate BoltzNet’s utility for bespoke biosensor design and to create a modeling framework applicable to other transcription factors, sensing applications, and organisms. This work has the potential to improve biosensor development, enabling more efficient and accessible diagnostics.