Despite their ubiquity, designing new biosensors remains difficult. Molecular recognition alone is not sufficient to create a functional sensor. A sensing protein must not only bind its target but also convert that interaction into a controlled response—one that is strong, rapid, reversible, and robust across diverse environments. At present, we lack general principles that explain how protein sequence encodes these sensing behaviors.

​

Our laboratory seeks to uncover these principles by systematically perturbing sensing proteins and measuring their responses at scale. Using high-throughput functional assays such as Sensor-seq, we screen large libraries of transcription factor variants against panels of candidate ligands. Each variant is linked to a molecular barcode, allowing thousands of sensor designs to be evaluated simultaneously in a single experiment. The resulting datasets map how sequence variation reshapes ligand recognition, signal propagation, and transcriptional output. 

By integrating these large-scale measurements with computational modeling and machine learning, we aim to learn general design rules for biological sensing materials. These models enable the rapid generation of new sensor designs for previously unseen molecules, dramatically accelerating the design–build–test cycle.
Ultimately, our goal is to create programmable biosensors—molecular systems that can be rapidly adapted to detect new targets, tuned for specific response behaviors, and deployed in applications ranging from field diagnostics to environmental monitoring and biosecurity