Fig. 4: MAGIK determines local diffusion properties.
From: Geometric deep learning reveals the spatiotemporal features of microscopic motion
a, Simulated single-object tracking experiment where fluorescence microscopy is used to follow the motion of single molecules performing Brownian motion with diffusivity D varying from particle to particle. Scale bar, 20 px. b,c, Ground-truth (b) and predicted (c) graphs. The edges depict the network of associations used to infer dynamic properties without direct linking. The nodes are colour-coded according to the value of the target feature, that is, the displacement scaling factor \(\sqrt{2D}\) measured in pixels per frame (colour bar in b). d, Probability distribution of the predicted versus ground-truth diffusion coefficient D, showing a good agreement (MAE = 0.06). e, Simulated single-object tracking experiment where fluorescence microscopy is used to follow the motion of single molecules performing Brownian motion with diffusivity D randomly varying in space. Scale bar, 20 px. f–h, Ground-truth (f) and predicted (g,h) diffusion maps. Ground-truth spatial diffusivity pattern (f) and prediction obtained by MAGIK using 100-frame-long (g) and 1,500-frame-long (h) videos with about 0.02 localizations per px2 per frame. The analysis is performed by breaking down the sequence into 2 and 30 videos of 50 frames each, respectively. Predicted maps are obtained by interpolating the values of diffusivity obtained for the nodes over the 64 px × 64 px grid through a triangulation-based nearest-neighbour algorithm.
