Extended Data Fig. 7: Quantitative comparison between the 2D and 3D analysis.
(a) Top panel: 3D SIM image sections were overlayed to produce a flattened image of the ER; left: central section of the ER image stack; right: colour coded flattened ER structures. Bottom: comparison of ERnet results for both types of images reveal nearly identical connectivity, supporting the notion that a 2D analysis suffices to capture ER topology in the cell types studied. Red spots: nodes; green lines: edges. Lower bottom: comparison of the connectivity metrics. To do this, we obtained volumetric sections of ER structures with SIM and reconstructed the data in 3D. We then applied ERnet on flattened image stacks or individual 2D sections and compared the two results. For phenotyping the effect of drugs in cell lines such as COS-7 or U2OS, which are morphologically flat, the advantages in speed and applicability of a 2D method greatly outweigh the disadvantage of not capturing the occasionally occurring 3-dimensional network features. N = 3 sectioning SIM images. Data are presented as mean ± SEM. See Source Data Extended Data Fig. 7a. (b) Same analysis to (a) performed on tubular domains only. N = 3 sectioning SIM images. Data are presented as mean ± SEM. See Source Data Extended Data Fig. 7b.
