And how is your search model related to your protein? Perhaps your model just needs substantial rebuilding. Are the maps after your semi-successful refinement job informative? Did you try tight jelly-body refinement with many (triple digits) cycles in Refmac? That sometimes gets you out of local minima.
Cheers,
Robbie
Sent with my Windows Phone
________________________________
Van: Dominika Borek<mailto:[log in to unmask]>
Verzonden: 23-10-2015 03:06
Aan: [log in to unmask]<mailto:[log in to unmask]>
Onderwerp: Re: [ccp4bb] Refinement doesn´t work
How do initial electron density maps (2Fo-Fc and Fo-Fc) look like?
D.
Aleksandar Bijelic wrote:
> Hi everybody,
> I am currently trying to solve a structure, which at the beginning
appeared to be "solveable". Processing went fine with the following
statistics (brief excerpt):
> 7.96 2144 741 763 97.1% 1.8% 2.0%
> 2091 49.44 2.1% 99.9* -13 0.721 287
> 5.67 3945 1214 1262 96.2% 2.9% 3.0%
> 3932 32.35 3.4% 99.9* -7 0.715 629
> 4.64 4631 1546 1605 96.3% 5.2% 4.7%
> 4537 23.75 6.3% 99.6* -10 0.795 561
> 4.03 6015 1839 1893 97.1% 6.6% 6.4%
> 5990 21.72 7.9% 99.5* -13 0.714 931
> 3.60 6760 2055 2120 96.9% 12.6%
> 11.9% 6733 14.11 15.1% 98.7* -11 0.750 1054
> 3.29 6596 2192 2308 95.0% 17.2%
> 17.2% 6436 7.56 21.0% 97.9* -8 0.716 843
> 3.05 8167 2464 2534 97.2% 21.6%
> 23.1% 8139 4.86 25.8% 98.1* -9 0.650 1318
> 2.85 8952 2627 2684 97.9% 34.4%
> 38.4% 8908 2.92 40.9% 97.1* -10 0.591 1592
> 2.69 8115 2535 2857 88.7% 50.7%
> 58.6% 7938 1.82 60.6% 93.7* -4 0.593 1284
> total 55325 17213 18026 95.5% 7.2% 7.4%
> 54704 12.92 8.7% 99.8* -9 0.673 8499
> Thus, the data looks/looked fine for a resolution of at least (roughly)
2.7 and I also MR worked fine with TFZ > 40 and LLG > 2000. However, I
failed to refine the structure since every refinement strategy I tried
increased Rfree constantly, that is, every single refinement step
increased my Rfree. I tried everything and I searched and found a lot of
refinement strategies and suggestions here in this board, but nothin
worked. I am asking myself what is wrong with this data, since Xtriage
didn´t find any pathologies. So, my Rfree starts from about 30-33% and
increases to 40-50% (depending on how many cycle I set, so every cycle
increases Rfree). I tried simulated annealing (both cartesian and
torsion) it didn´t work. I tried rigid body refinement ("desperation
move") and Rfree increased. The "best" refinement increased my Rfree
from 31 to 32% (coordinates, NCS, TLS and group ADP with group selection
(with and without applying rigid body)). I have no idea what the problem
is. In most cases Rwork decreases but sometimes it also increases. I
hope somebody can help me! Maybe I have overseen something? Maybe the
data was processed wrongly? Maybe I have introduced wrong adjustments
(accidentally)? Maybe someone can tell my what to try or suggest me a
refinement strategy (maybe I am doing soemthing terribly wrong, but also
"default" refinement increased Rfree dramatically). Thanks in advance!
Please ask me for more information if needed!
> Cheers,
> Aleks
Dominika Borek, Ph.D. *** UT Southwestern Medical Center
5323 Harry Hines Blvd. *** Dallas, TX 75390-8816
214-645-9577 (phone) *** 214-645-6353 (fax)
