Directed Evolution of Near-Infrared Serotonin Nanosensors with Machine Learning-Based Screening.

In this study, we employed a novel approach to improve the serotonin-responsive ssDNA-wrapped single-walled carbon nanotube (ssDNA-SWCNT) nanosensors, combining directed evolution and machine learning-based prediction. Our iterative optimization process is aimed at the sensitivity and selectivity of ssDNA-SWCNT nanosensors. In the three rounds for higher serotonin sensitivity, we substantially improved sensitivity, achieving a remarkable 2.5-fold enhancement in fluorescence response compared to the original sequence. Following this, we directed our efforts towards selectivity for serotonin over dopamine in the two rounds. Despite the structural similarity between these neurotransmitters, we achieved a 1.6-fold increase in selectivity. This innovative methodology, offering high-throughput screening of mutated sequences, marks a significant advancement in biosensor development. The top-performing nanosensors, N2-1 (sensitivity) and L1-14 (selectivity) present promising reference sequences for future studies involving serotonin detection.