AI-driven photophysics-aware design of fluorescent probes with applications in α-synuclein biosensing and inhibitor screening.

Target-specific fluorescent probes are essential for biosensing and drug screening, yet their rational design remains challenging due to the intrinsic trade-off between target binding and photophysical properties. Generative artificial intelligence offers new opportunities for molecular design. However, its application to probe design remains limited as photophysical properties are not encoded during generation and resulting molecules are often chemically implausible or synthetically inaccessible. Here we present a unified generative framework in which photophysical evaluation is embedded into structure-aware molecular generation, enabling simultaneous optimization of target recognition and optical functionality during molecular construction. Furthermore, we establish a consensus-driven prioritization strategy that incorporates physics- and synthesis-constrained fragment growth, ensuring the chemical realism and synthetic accessibility of the generated molecules. As a proof-of-concept, we applied this strategy to α-synuclein fibrils, a key pathological hallmark of Parkinson's disease, and identified a fluorescent probe (CLead1) with a strong turn-on response and high selectivity against competing amyloid species. Leveraging CLead1, we developed a high-throughput screening assay and identified Dryocrassin ABBA as a neuroprotective inhibitor of α-synuclein fibrillization. This work establishes a broadly applicable generative strategy for designing target-specific fluorescent probes, bridging biosensing and therapeutic exploration.