I’m new to cryoem and new to cryospock, so I’m struggling with even the basic steps, but I’m getting a lot of help from the “cryospock discussion”.
As a result of all the refinements, two versions of the map are generated: ‘map’ and ‘map sharp’ are generated, and my structure is more visible in the map than in the map sharp and does not look overwhelming when viewed in CHIMERA. Is it up to the researcher to choose between map and map sharp?
Due to various technical reasons, a cryo-EM reconstruction has a power spectrum that decays with resolution instead of becoming flat (‘white’) as it should for a real protein structure. This decay is roughly log-linear so we can look at the log power spectrum after some resolution where it is most linear (like 10Å), and fit a slope to it. You can see this process in the Guinier analysis plots from the refinement. We call the estimated slope the “B-factor” and use it to correct the power spectrum by boosting the high frequency information. This approach is the most commonly used “sharpening” method in cryo-EM today.
While boosting the high frequency information like this does make the cryo-EM density more like a real protein structure, it also boosts up noise (which is always stronger at higher frequency/resolution). This creates the ‘dust’ in the map you see outside and around the protein itself. Furthermore, cryo-EM reconstructions usually have varying resolution throughout the map. Each region of the map thus has its own correct B-factor. When the global B-factor is too high for a local region of the map, you will also see continuous densities get chopped up, stretched out, etc. - all effects from having spuriously high power at high spatial frequency.
A rule of thumb is that if the B-factor is less than ~150 and the resolution is better than 4 Å, the sharpened map is probably OK. Especially for a very high resolution structure, with very uniform resolution, the sharpened map will be best for model building. However, in practice we must always look at both the raw map and the sharpened map, and be careful not to over interpret the fine densities we see. There are also alternative sharpening methods, like DeepEMhancer, LocScale, and so forth. In any case, we should always check the raw map and be careful with what we think are real features.
If the B-factor is very high (above ~300) or the resolution is worse than 4.5Å I would not bother looking at the sharpened map, it will probably be too “chopped up” or “blown out” or “over sharpened” to be useful.
Thanks for the detailed and easy to understand explanation.
I will follow your advice and always compare the raw map and sharpened map.
Thank you so much!