Rapid, Accurate Classification of Single Emitters in Various Conditions and Environments for Blinking-Based Multiplexing

Although single-molecule imaging is widely applied in biology and materials science, most studies are limited by their reliance on spectrally distinct fluorescent probes. We recently introduced blinking-based multiplexing (BBM), a simple approach to differentiate spectrally overlapped single emitters based solely on their intrinsic blinking dynamics. The original proof-of-concept study implemented two methods for emitter classification: an empirically derived metric and a deep learning algorithm, both of which have significant drawbacks. Here, a multinomial logistic regression (LR) classification is applied to rhodamine 6G (R6G) and CdSe/ZnS quantum dots (QDs) in various experimental conditions (i.e., excitation power and bin time) and environments (i.e., glass versus polymer). We demonstrate that LR analysis is rapid and generalizable, and classification accuracies of 95% are routinely observed, even within a complex polymer environment where multiple factors contribute to blinking heterogeneity. In doing so, this study (1) reveals the experimental conditions (i.e., P_exc = 1.2 μW and t_bin = 10 ms) that optimize BBM for QD and R6G and (2) demonstrates that BBM via multinomial LR can accurately classify both emitter and environment, opening the door to new opportunities in single-molecule imaging.