Fig. 6

Position-dependent weak binding detected by nonparametric Bayesian inference. a A representative trajectory analyzed by nonparametric Bayesian inference. The blue line indicates the trajectory of lever-arm-less S1 labeled with GNP. Magenta and cyan plots indicate the inferred mean position of the transient binding states and detached state, respectively. Data were obtained from three independent and reproduced experiments. Representative data are shown. b An expanded trace of the box in a. c A histogram of the detected transient binding position for 1,000 inferences (see Methods). Positions were fit to five Gaussian distributions with peaks of −26.0 ± 0.8, −12.9 ± 0.8, −1.1 ± 0.7, 9.9 ± 0.7, and 22.8 ± 1.2 nm, respectively. The calculated accessibilities (access ratio) are 0.014 ± 0.004, 0.184 ± 0.018, 0.493 ± 0.022, 0.245 ± 0.021, and 0.064 ± 0.009, and the dwell times are 322 ± 87, 237 ± 37, 160 ± 11, 238 ± 22, and 513 ± 61 μs. These values are shown in Fig. 7a except for the leftmost peak indicated by the dashed line (position: −26.0 nm, access ratio: 0.014, dwell time: 322 μs), because the accessibility is too low to infer the dwell time. n = 4660. Errors in the values indicate standard deviations. d–g Dwell times of inferred transient states for a typical inference analysis. Dwell times were fit to the cumulative distribution function of a single exponential decay. The obtained dwell times were 299 ± 25 (d), 160 ± 5 (e), 278 ± 19 (f), and 486 ± 57 μs (g). The corresponding inferred positions were −12.5, −1.3, 10.0, and 21.9 nm, with n = 267, 708, 328, and 85, respectively.