Single molecular tracking is a valuable approach for investigating the dynamic processes and molecular interactions in soft matter systems, particularly in biological systems. However, understanding the complexity of single molecule motion behavior in biological systems remains a significant challenge. To address this issue, we propose a two-step classification method based on unsupervised learning to efficiently identify and classify single molecule trajectories. Firstly, we employ an entropyconstrained least squares method to differentiate between confined (e.g., immobile) and unconfined diffusion trajectories. Subsequently, statistical tests are utilized to categorize the unconfined trajectories into different diffusion modes such as sub-diffusion, normal diffusion, and superdiffusion (Figure 1).
By applying this methodology, we analyze the diffusion motion of single molecules in both DOPC model cell membranes and living cell membranes while uncovering their distinct responses to cholesterol composition. Our findings demonstrate that both model membranes and living cell membranes exhibit diverse molecular diffusion modes. Specifically, in the DOPC model membrane system, the presence of cholesterol components impedes lipid diffusion within the membrane. The degree of inhibition is positively correlated with the amount of cholesterol present. For instance, as the cholesterol content in the membrane increased from 0 to 20% (DOPC:Chol = 4:1) and 50% (DOPC:Chol = 1:1), there was an increase in the proportion of molecules exhibiting confined diffusion and sub-diffusion (from 55% to 45%), while there was a decrease in the proportion of molecules displaying normal diffusion and super-diffusion (from 45% to 35%). The ensemble diffusion coefficient of molecules in the membrane was significantly reduced, which can be attributed to both a decrease in velocity among fast-moving molecules (indicated by downward shift of peak A in Figure 2a, marked by red arrows) and an increase in slow-moving molecules (indicated by upward shift of peak B in Figure 2a3). Interestingly, upon removal of cholesterol using MeβCD, single-molecule mobility within the DOPC/Chol composite membrane systems was restored to levels similar to that observed for pure DOPC membranes (blue arrows in Figure 2a).
Conversely, in the live cell membrane system, the diffusion coefficient values of molecules are significantly lower compared to those observed in the model membrane system; furthermore, removal of cholesterol further decelerates molecular diffusion rates (demonstrated with orange arrows in Figure 2b). This study contributes towards understanding the intricacies of biomolecular motility and its dependence on environmental factors from a perspective of single molecular motion.