Mesh Generation

[KTDownes]

Hello! - Big fan of this work :)

I have been playing with the extrapolation distance with my own data - intracellular vesicles with their tops and bottoms cut off by the missing wedge. I am struggling to find the sweet spot for mesh generation. Extrapolation values of 0.512 - 1.5 nm result in complete closure of the vesicles, despite a large section missing in the segmentation. However a 0.51 nm extrapolation distance results in under sampling of the membrane with multiple surfaces dotted over the membrane rather than joined. I'm running values between 0.51-0.512nm now in a trial and error manner, but I was wondering whether anyone could offer me any help or advice from their experience? Or provide some information for the rationale behind a "good" extrapolation distance?

I've attached my a screenshot of my segmented structure and the over-extrapolated and under-extrapolated results.

*editing this to update the extrapolation values I have tried

[bbarad]

Hi, sorry for the long delay on responding to this! Something seems very wrong with the extrapolation values here - are you working in Å mode by any chance? The extrapolation distance will be in nanometers or angstroms depending on which mode you use. You definitely can't drop your extrapolation distance under the pixel size. If you'd be willing to share an example tomogram, I can have a look at why its closing so aggressively, but my best guess is that the surface cleaning is not working properly for some reason.

Generally, a "Good" extrapolation distance should be ~1-1.5 the (binned) voxel size or larger (if you want to more aggressively extrapolate!). You should not be getting closure with a 1nm gap here, it seems like a totally reasonable set of settings.