Extended Data Fig. 2: Nanodomain association of model LD cargo analyzed using Gaussian KDE.
From: Membrane bridges and nanodomain partitioning govern membrane protein targeting to lipid droplets
a. A Gaussian mixture model was applied to the Dapp distributions of GPAT4 (left) and HSD17B13 (right) to resolve distinct mobility states within each population. Quantile–quantile (Q–Q) plots and the corresponding Dapp distributions of the fast (purple) and slow (orange) mobility populations are shown. Mixing weights, means and variance are written on the plots. b. Representative examples of GPAT4 tracks localized to LDs (top) and the ER (bottom), showing full-track KDE density maps (left). Tracks were segmented into 100 ms windows, and regions exceeding the dataset’s median KDE density were defined as ‘dense.’ The area of these dense regions was quantified. c. Apparent diffusion coefficients were calculated for HSD17B13 tracks using a 5-ms rolling window and clustered by whether segments were inside (orange) or outside (purple) nanodomain regions. Median values are indicated by dotted lines. Nsteps = 502,787 (LD), 1,173,290 (ER). A two-sided Mann–Whitney test was used to assess the statistical significance of the difference of Dapp inside and outside the nanodomains. P = 4.4 × 10−215. d. Distributions of the fast and slow mobility populations were derived using a Gaussian mixture model from Extended Data Fig. 2a and classified based on whether segments fell inside (left) or outside (right) the nanodomain boundaries. The slow-moving population is predominantly enriched within nanodomains, whereas the fast-moving population is the major species outside these nano-regions. Median Dapp values of each population are stated on each graph. Mixing weights πinside = [0.71 (slow), 0.29 (fast)], πoutside = [0.27 (slow), 0.73 (fast)]. σ2inside = [0.001 (slow), 0.005 (fast)], σ2outside = [0.001 (slow), 0.0068 (fast)]. e. Mean KDE density within dense regions was computed by dividing the summed KDE values within dense regions by the number of localizations per segment, for ER (purple) and LD (orange) tracks. The box represents the 25th–75th percentiles (IQR), the center line indicates the median. A two-sided unpaired t-tests was used analyze significant differences: P = 1.15 × 10−18 [NER = 609, NLD = 823] for GPAT4, 1.88 × 10−83 [NER = 3036, NLD = 1332] for HSD17B13. f. Total KDE density per segment was normalized to the grid area and compared between ER (purple) and LD (orange) tracks. The box represents the 25th–75th percentiles (IQR), the center line indicates the median. A two-sided unpaired t-tests was used analyze significant differences: P = 1.57 × 10−17 [NER = 609, NLD = 823] for GPAT4, 1.21 × 10⁻65 [NER = 3036, NLD = 1332] for HSD17B13. g. Density enrichment, calculated as the ratio of mean KDE density within dense regions to total normalized KDE density per segment, is shown for ER (purple) and LD (orange) tracks. Source numerical data are available in source data.
