Fig. 4.

Diffusional fingerprinting applied to three widely diverse systems. A, D, and G display the one-dimensional LDA projection showing a threshold that maximizes the sum of sensitivity (proportion of correctly identified positives) and specificity (proportion of correctly identified negatives). B, E, and H display the features ranked by their components in the LDA projection. C, F, and I display 30 exemplary traces from the respective dataset used to construct B, E, and I. (A–C) Diffusional fingerprinting of the TLL L3 variant measured on two different substrates, lard and trimyristin (n = 6,270). (D–F) Diffusional fingerprinting of the two halo-tagged transcription factors Sox2 in mouse embryonic stem cells and NLS in human U2OS cells from a study by Hansen et al. (58) (n = 8,937). (G–I) Diffusional fingerprinting of differently coated nanoparticles diffusing in mucus on top of a lipid membrane (n = 2,286). The confusion matrix from a fivefold stratified cross-validation of a logistic regression classifier for each dataset is shown in SI Appendix, Fig. S13.