Dear Lingyun,
The auto-refine calculates resolution using gold-standard FSC curves on
unmasked maps. Sometimes, for very small proteins this may result in an
under-estimation of resolution that is just too much. What kind of
orientational accuracies do you get?
One can use a single-reference 3D classification procedure to refine
structures as well (that's how we worked in relion-1.0). By continuing
old runs one can use increasing angular sampling rates. However,
overfitting may be a real danger here! Be very careful not to put the
regularisation parameter (T) too high. One could perform high-resolution
noise substitution (Shaoxia Chen et al, Ultramiscroscopy, 2013, 135,
24-35) to be sure about the level of overfitting.
HTH,
S
On 05/26/2014 12:14 PM, Lingyun Zhao wrote:
> Hi, Sjors,
> I am a graduate student in Tsinghua University. Now I am using RELION to do single particle reconstruction. My project is about a small protein complex, no more than 300 KD and has no symmetry. I did all steps in RELION and all the results looked good, except for the auto refine result. Auto refine made the results much worse than before and the map looked ugly. So what may lead to this happen? What is your suggestions about this? Thank you!
>
>
>
>
> Best,
> Lingyun
--
Sjors Scheres
MRC Laboratory of Molecular Biology
Francis Crick Avenue, Cambridge Biomedical Campus
Cambridge CB2 0QH, U.K.
tel: +44 (0)1223 267061
http://www2.mrc-lmb.cam.ac.uk/groups/scheres
|