Bayesian analysis and efficient algorithms for single-molecule fluorescence data and step counting

With the growing adoption of single-molecule fluorescence experiments, there is an increasing demand for efficient statistical methodologies and accurate analysis of the acquired measurements. Existing analysis frameworks, such as those that use kinetic models, often rely on strong assumptions on the dynamics of the molecules and fluorophores under study that render them inappropriate for general purpose step counting applications, especially when the systems of study exhibit uncharacterized dynamics. Here, we propose a novel Bayesian nonparametric framework to analyze singlemolecule fluorescence data that is kinetic model independent. For the evaluation of our methods, we develop four MCMC samplers, ranging from elemental to highly sophisticated, and demonstrate that the added complexity is essential for accurate data analysis. We apply our methods to experimental data obtained from TIRF photobleaching assays of the EphA2 receptor tagged with GFP. In addition, we validate our approach with synthetic data mimicking realistic conditions and demonstrate its ability to recover ground truth under high- and low-signal-to-noise data, establishing it as a versatile tool for fluorescence data analysis. Significance statementProtein complexes are critical for cell function. Advances in fluorescence experiments have facilitated their direct study, achieving single-molecule resolution in the determination of their stoichiometry. However, the analysis of raw fluorescence data remains a challenge. Traditional visual inspection methods are often time-consuming and susceptible to user bias, while conventional statistical approaches are limited by strong assumptions regarding molecular and probe dynamics or exceedingly high computational requirements. In this study, we introduce a novel statistical methodology for robust and efficient analysis of fluorescence data. Our innovative approach not only enhances the accuracy and reliability of data analysis following a fluorescence experiment but also allows high-throughput applications. Our methods present a significant advancement that may extend the scope of current fluorescence techniques.