LAUSR

Cellular signaling is regulated by the spatiotemporal dynamics and kinetics of molecular behavior. To investigate the mechanisms at the molecular level, fluorescence single-molecule analysis is an effective method owing to the direct observation of individual molecules in situ in cells and the results in quantitative information about the behavior. The integration of machine learning into this analysis modality enables the acquisition of behavioral features at all time points of all molecules. As a case study, we described a hidden Markov model-based approach to infer the molecular states of mobility and clustering for epidermal growth factor receptor (EGFR) along a single-molecule trajectory. We reveal a scheme of the receptor signaling through the dynamic coupling of the mobility and clustering states under the influence of a local membrane structure. As the activation process progressed, EGFR generally converged to an immobile cluster. This state exhibited high affinity with a specific cytoplasmic protein, shown by two-color single-molecule analysis, and could be a platform for downstream signaling. The method was effective for elucidating the biophysical mechanisms of signaling regulation when comprehensive analysis is possible for a huge number and multiple molecular species in the signaling pathway. Thus, a fully automated system for single-molecule analysis, in which indispensable expertise was replicated using artificial intelligence, has been developed to enable in-cell large-scale analysis. This system opens new single-molecule approaches for pharmacological applications as well as the basic sciences.