When you have some reasonable map features but such a high R factor my first guess is you have the spce group wrong - P21212 instead of P212121 or something. It can mean you have half the reflections corret and half not.
Go back and check the data processing carefully (CCP4 GUI2 gives an excellent report of what you should worry about! Is there a non-crystallographic translation vector? If you have a translation (x,y,1/2] say then the absences along the l axis may not indicate a screw axis. Is it an enantiomorphic SG such as P31 or p32 - ones you cannot select before solving the structure.. Agree with Dale - R factors of 4% cannot generate a flat map.. Eleanor On 21 January 2017 at 07:56, Dale Tronrud <[email protected]> wrote: > The R value is basically a sum of all the Fourier coefficients of the > difference map relative to their Fobs amplitude. If your R value is in > the 40% range you MUST have a difference map with a lot of stuff in it. > The only way you can have a clean difference map with an R value so > large is if there is a mistake either in the calculation of the R value > or the map. > > I think you need to double-check your work. > > Dale Tronrud > > P.S. Are you contouring your difference map at a reasonable level? > Think e/A^3 not "sigma"s. When I look at a difference map I start at a > level of 0.18 e/A^3 and then think about other levels. There is nothing > magical about that number but looking at all maps the same way helps you > lean what a map should look like. If you want an internal calibration, > you can leave out one, well ordered, water molecule. Then you can see > what the difference density due to an absent atom looks like in YOUR > map, and compare that to the other features you see. > > Remember, if you have stuff EVERYWHERE in your difference map, you > will see NOTHING at a 3 rmsd contour because the rmsd of the map is huge! > > On 1/20/2017 10:45 PM, ameya benz wrote: > > I am trying to refine my data set collected at 1.9A. The density appears > > clean and the fit is also good. In coot, all residues are in green zone > > in density fit analysis. Also in Ramachandran plot 89% residues are in > > allowed region. Few residues in loop region do not have good density for > > side chains. My protein contains about 61% loops and remaining beta > > sheets , no alpha helices (predicted from homology model). What could be > > done to improve the Rfree and Rfactor? >
