Figure 4.
Predictive ER model resolves the structural complexity of the ER network during different stages of the cell cycle. (A) Representative examples of ER predictions in naïve cell 15 (SVG-A) processed during interphase as described in Fig. 3 showing the characteristic network of ER sheets connected at branch points to ER tubules. ER tubules were more abundant toward the periphery of this cell and ER sheets were more abundant closer to the nucleus. For clarity, manual VAST editing was used to eliminate pixels of false positive predictions associated with the nuclear envelope. Scale bar, 1 μm. (B) Representative examples of ER predictions from a mitotic naïve cell 8 (SUM 159) prepared by CF and imaged isotropically at 10 nm; the ER model was trained with ER ground truth annotations from interphase cells 1 and 2 (HEK293A) prepared by CF visualized isotropically at 5 nm resolution and downsampled to 10 nm. It shows successful recognition of an extensive network of fenestrated ER sheets (red arrow heads) connected to ER tubules, characteristic of mitotic cells. Ground truth annotations used to train the interphase ER model did not contain ER fenestrations, as they are barely present during stage of the cell cycle. Darker regions corresponding to chromosomes are outlined with yellow dotted lines. Scale bar, 3 μm (see Video 3).

Predictive ER model resolves the structural complexity of the ER network during different stages of the cell cycle. (A) Representative examples of ER predictions in naïve cell 15 (SVG-A) processed during interphase as described in Fig. 3 showing the characteristic network of ER sheets connected at branch points to ER tubules. ER tubules were more abundant toward the periphery of this cell and ER sheets were more abundant closer to the nucleus. For clarity, manual VAST editing was used to eliminate pixels of false positive predictions associated with the nuclear envelope. Scale bar, 1 μm. (B) Representative examples of ER predictions from a mitotic naïve cell 8 (SUM 159) prepared by CF and imaged isotropically at 10 nm; the ER model was trained with ER ground truth annotations from interphase cells 1 and 2 (HEK293A) prepared by CF visualized isotropically at 5 nm resolution and downsampled to 10 nm. It shows successful recognition of an extensive network of fenestrated ER sheets (red arrow heads) connected to ER tubules, characteristic of mitotic cells. Ground truth annotations used to train the interphase ER model did not contain ER fenestrations, as they are barely present during stage of the cell cycle. Darker regions corresponding to chromosomes are outlined with yellow dotted lines. Scale bar, 3 μm (see Video 3).

This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal