Hello everyone!
I am trying to understand the significance of the Real-Space Slices graph from my NU-refinement. I understand that the heat map is meant to represent the scalar density value, but how do I use this information? I have heard that more blue in the plot is preferred over red, but why?
I am master’s student trying to gain a good understanding for my thesis, and all input is appreciated! Thank you in advance 
Hello CryoCrier,
Those heat maps are very useful for a quick inspection of the results, with no need to open the 3D volumes to make rough comparisons. The rainbow goes from blue to red, blue beeing the lowest density, red being the highest. Remember those are projections, so thicker parts of the structure will appear warmer. What you’ve heard probably means you need to achieve a blue background, which means your map will be less noisy at the end. If your background is appearing red, then definitely there is something really wrong. You can open the map in e.g. Chimera and try to align it with the three projections, one at a time, as an exercise. Use the threshold slide in Chimera to compare results with different background colors. Warmer backgrounds in the projections view will appear at a higher threshold than the ones with cold backgrounds.
The slices in NU-refine (and other job types involving 3D reconstruction) are just that - central orthogonal slices of the map, not projections (sometimes projections would be useful!).
Perhaps a little confusingly though, the thumbnails shown in the job cards and the volume outputs are projections, not slices.
All right, I see. Thanks!
In addition to the good notes above, I just want to add one more point — the color scale always expands to cover the values in the slice. This means that if there is a region of your map that is slightly negative (not uncommon), a good background (relativley flat, near zero) will not necessarily be blue, but can be green/yellow (e.g., this 2.8 Å map from EMPIAR 11631)
Generally, when I look at these slices, I look for two things:
Here’s an example of a refinement with far too small a mask. You can see that where my “particle” is, the slice is still basically the same color as the background: there’s nothing there! This refinement claims to be ~4Å, but I can tell just by looking at this slice that that’s overfitting to a too-small mask.
To piggyback on this post – can you give me some feedback on this real space slice? I am trying to work on a relatively flexible protein that is part of a protein complex. In this specific refinement (a local refinement), I had carried out the local refinement after doing symmetry expansion + 3D classification with the best classes from the 3D classification. I had been running local refinements where I increase the max alignment resolution and the following real space slices came up, and I have been concerned about if there is something I can do to improve my alignments. It notes that this is capped at ~5.5 A, but I don’t know if I can believe it given what you noted with the latest example – I should also clarify that the C like shape in my slice is exactly the shape that I expect from the region I am trying to locally refine.
Hi @gmperez!
These slices are meant to make it easier to tell what is going on with a refinement while it’s running without downloading the map. It’s always going to be difficult (or perhaps impossible) to figure out why a refinement may be failing just from 2D slices, so you should always download the map and inspect it there when trying to diagnose any failures.
That being said, it sounds like you’re trying Local Refinement on a relatively small region of the overall protein. If that’s the case, it can be challenging to prevent overfitting. You may want to try a larger mask, or an alternative technique like 3DVA that can handle continuous heterogeneity without the need for such a small mask.
If you have any more questions, please feel free to ask them!
