Artificial allosteric protein switches with machine-learning-designed receptors.

Machine-learning-designed minimal ligand-binding domains enable efficient allosteric switches without global conformational change, facilitating diverse biosensors and synthetic systems.

• Why it matters: Understanding and engineering artificial allosteric proteins is crucial for advancing synthetic biology, energy processing, and biotechnology applications, yet designing effective switches remains challenging.
• What they did: The study used machine learning to create minimal ligand-binding domains that function as receptors in single-component allosteric switches, forming colorimetric, luminescent, and electrochemical biosensors, and constructing fully synthetic switches with artificial receptor and reporter domains.
• The result: Ligand binding reduces conformational entropy, boosting reporter activity, enabling practical applications such as steroid-dependent antibiotic resistance in E. coli and bioelectronic devices for hormone quantification.