First question - is the original PDB structure twinned? Can you download
the data and check that..
And you have generated the extra models correctly have you - there are
origin shifts to consider I think..
2) How can you get an R factor of 20.7 / 25.7 with REFMAC in P6122 and
one of 27.77/36.2 in SHELX for the same spacegroup?
3) HKLF 5 is not needed for merehedral twinning - the lattice for the
two twin parts overlap exactly here
4) P3112 k,h,-l 0.177 0.22 0.494 and
P3112 -h,-k,l 0.177 0.22 0.494
are exactly equivalent - the two operators are symmetry equivalents ..
And ditto for
P3121 -k,-h,-l 0.177 0.213 0.494
P3121 -h,-k,l 0.177 0.213 0.494
But none of that explains why you get the same result for different
space groups - I suppose it could be twinned in more than one way -
cant remember the technical term for it..
Kerff Frédéric wrote:
Hello,
I'm refining a structure already available in the pdb with a ligant. The
apparent SG (P6122) and the unit cell parameters are identical. Without
taking into account any twinning, I get a R and Rfree values of 20.7 and
25.7 at 1.75A with refmac5. I agree these are very decent values but I
unfortunately thought I could do better. I double checked the data.... and
all the different criteria agree, there is some twinning...
So I switched to shelx and tried few runs (without the ligant because it's
not recognize) to find the kind of twinning I have:
R Rfree Fract
P6122 0.277 0.362
P61 0.266 0.35
P61 twin 0.177 0.219 0.494
P3112 0.27 0.352
P3112 k,h,-l 0.177 0.22 0.494
P3112 -h,-k,l 0.177 0.22 0.494
P3121 0.27 0.357
P3121 -k,-h,-l 0.177 0.213 0.494
P3121 -h,-k,l 0.177 0.213 0.494
It seems I have for second time in a year a tetartohedral twinning in the
space group P31.
I didn't process it the right way yet (I know it's with HKLF 5 and a
modified .hkl file) but I compared some omit map from Refmac5 and shelx in
Coot anyway:
- with refmac I see a nice and clean density corresponding to my ligant in
both the difference and the 2fo-fc maps
- with shelx things are very noisy in the ligant area and I have to use a
very low cutoff to see anything. I even got a good density around a
completely messed up ligant. In other words shelx maps seem very biased by
the twinning.
What should I do?
- Use Refmac5 without twinning because the maps are not biased and the
statistics very decent (i'd love everybody telling me so)
- or how do I take twinning into account without getting biased maps?
Thank you for reading this until the end and for any useful suggestion.
Fred
