I am using Fo-Fc, 2Fo-Fc, and a DM maps to build my protein in COOT at ~3A resolution. I know the gene sequence of my protein. What is the ideal contour level(sigma) setting for building into a Fo-Fc map? I have read somewhere it should be ~2.5. But WHY 2.5? Why not 1.0 sigma like the 2Fo-Fc map contour setting recommendation?. I also have a very important part of my protein that has poor 2Fo-Fc and DM density in this region, but has some good density in this region at ~1.0-1.5 sigma for the Fo-Fc map.
Do you have any NCS?
If you are using NCS averaged maps and the NCS molecules differ where you have weak density, then you need to change your NCS parameters.
You can Just build carefully in small stages at lower sigma (~1.5), and make regular refinement and map calculations after small ammounts of building, checking the Fo-Fc maps at each stage for negative density. A bit tedious but thats life!
Depending on how much you have already built, perhaps a composite omit map might look better and be easier to build into?
As a last resort, if you can't build it there are also cases where highly redundant datasets, or soaking crystals with solvents have been used to improve the density for disordered protein regions.
As with everything crystallographic there is always a million things to try!, good luck.
Think about what the number means. Signal to noise. Anything above 1.0 contains some signal. Anything above 15 sigma (or insert you favorite number here) is a strong signal. It is a matter of convincing yourself, your referees and your readers that (a) there is something there and that (b) you have identified/modeled it correctly.
So use all the tricks available to build it (low contours, NCS, composite omit maps, better data set, biochemistry experiments), and then be very strict with quality control: is there any reason for this compound to be there (e.g. there is 1 Molar of it in the solution), is the map improving (taking into account the resolution - 3 A is not that high), is the Rfree going down (or at least not going up), is the geometry good, is the chemistry making sense etc etc.
And just be honest about it - say you tentatively tried to model compound xxx in very weak density and explain which elements are in favor and which elements are against your interpretation of the electron density map.
Remember that there will always be density you can't model. Also, have a look at a very high resolution map (0.9 A or so) when you get the chance - there is a lot happening with the protein and the waters that you can't see at 3.0 A (alternate conformations, "split" waters, etc.).
And finally: put a deadline on it. If you can't model it in a few days, let it go.
1.5 sigma is really low for a difference map. There is a fair chance that you are below the noise level. You can find a good sigma level for the difference map is by starting high (say at five) and then scrolling down to the level at which nonsense negative difference density appears. A more elaborate way of finding the noise level is looking at the sigma value where the positive and negative peaks are roughly equal in amount.
Keep in mind that positive difference density without signal in the 2mFo-DFc map is suspicious once you are past the early stages of model building.
Thank you everybody for the insightful input. Helps me better understand that "noise" threshold is relative, in a sense. to the molecule you are working with. I hate to accept some "cut-off" value without understanding WHY. I forgot about using omit and NCS average maps, that helps.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Atomic, Molecular and Optical Physics
- Atoms