Computationally Designed Nanobinders as Affinity Ligands in Diagnostic and Therapeutic Applications

Detecting protein biomarkers is critical in fundamental research and clinical investigations of extracellular vesicles (EVs). Despite the prevalent use of antibodies as recognition elements, their application is often limited by challenges such as cross-reactivity and inconsistent affinity. Here, we report designed nanobinders (DNBs) as a promising alternative for EV protein analysis. Our approach adopted recent advances in de novo protein design driven by machine learning. We specifically developed a computational pipeline for optimizing DNB design, complemented by a robust validation in vitro. As a proof-of-concept, we engineered a PD-L1-targeting DNB. It demonstrated superior performance compared to antibodies, exhibiting up to a 51-fold increase in signal intensity during cellular imaging and enhanced sensitivity and selectivity in EV analysis. Moreover, the PD-L1 DNB was effective in inhibiting immune checkpoints. These findings underline DNB's potential as a reliable and scalable platform for EV-based diagnostics and therapeutics.