In coot, select. Validate > Difference Map Peaks > Find Peaks Above 4 r.m.s.d > [Find Peaks]
What do you see? Always remember, the real-space representation of your Rwork is the Fo-Fc map. The biggest peak or valley in this map is usually dominating your Rwork/Rfree. If you feel your Rwork/Rfree are too high, it is best to think of something sensible to put into your biggest Fo-Fc peak. If you have a big, negative feature in the middle of nowhere, try adjusting your bulk solvent parameters in refinement. It is also a good idea to look at your data quality stats. If I/sig(I) is 5, then I wouldn't expect Rwork/Rfree to ever drop below ~20%. This is because the error in the data is already about 20%. If your model fits better than that, it is probably modelling noise. If you have no difference features to speak of and your Rwork/Rfree are still high, it can be a good thing to prune back your model to include only the atoms you are really sure about. That is, well-ordered main chain, and maybe a few strong side chains like disulfides and obvious aromatics. Refine this heavily pruned-model to convergence. Don't worry about the R factors being high, they are supposed to be high when large parts of the model are missing. By "convergence" I mean until the atoms stop moving. R factors leveling off is not "convergence". If your atoms are still wandering about and changing B factors, then run the refinement one more time. Wait for things to settle. This is kind of like running a column, you want to wait until everything has equilibrated before you inject something new. This kind of equilibration is the only way to know that the rise or fall in R you see is due to the change you just made, as opposed to an after-effect of something you did a few dozen cycles ago. Once that is done, look at the top feature in your Fo-Fc difference map. This tallest peak is the least likely thing in your unit cell to be wrong. Build in this feature. Then re-refine to convergence again. This can take a while, but it is the best way to ensure that you avoid model bias of any kind. Building just one feature at a time is the most conservative strategy. If you're in more of a hurry, you might consider anything above 6 sigmas to be "safe", but 5 is pushing it, and 4 is dangerous unless there is nothing else left in the map. Happy Building! -James Holton MAD Scientist On 3/23/2019 9:17 PM, StrBio wrote: ALL. I have data at 2.4 A in P21 sp gr, helical protein. Refined to Rwork 29 Rfree 34 with nice density map and all nice statistics oither Rfactor (by Phenix). Refmac quit same. Should I deposit it or look better data? Any suggestion? ________________________________ To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1 ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
